The previous week, I had migrated the code base from XR interaction SDK to Oculus Integration SDK for Hand Based tracking and gesture control.

The oculus SDK caused the existing UI un-interact-able with ray interactor. Because most of the UI is done using canvas, I had to revert back to XR Integration SDK.

After that,  I created the Audio Record Button for Recording the audio using the inbuilt microphone in Quest 2. The audio was clear and had no issues during playback.

The Messages sent are rendered in a Button. And when we press the button or chat we play the audio attached to it.

The existing plan is to create a flask server where audio can be stored and the users fetch using chat ID.

That's all for this week

Bye!!

This week I tried migrating from XR Interaction toolkit to Oculus Integration SDK as Oculus SDK provided Hand Based tracking which would be useful for Gesture based control.

Script for Basic Audio Recording using Mic has also been completed.

The Oculus SDK makes use of OpenVR backend, this posed an issue as OpenVR is still in experimental stages.

The UI based control for Microphone is still pending. This would entail recording and sending or broadcasting to all connected clients

That's all for this week

Ciao

This week I worked on implementing the UI and UX for the Chat Module in VR. The UI was designed using Canvas in Unity.

The Canvas was then attached XR Rig Camera. It was converted that is rendered in world space mode.

A graphic renderer was added to the canvas element in order to interact with the XR Ray interactor.

As of now, in order to test the connection, a mock android app was also designed for testing inter device and client server communication

That's all for this week

Adios

This week I worked on creating the chat portion of the application. I made use of TCP and sockets to design a Mock Android App that would work like a Multi client setup.

Here, the Application is able to connect as well start a server. In the usual setup, the teacher would start the server and students would join to it.

Initial road blocks in implementation in VR is regarding the UI elements not rendering on screen. The sockets part remains untested and will be tested in the forthcoming weeks

That's all for this week!

This week was all about designing the static template for the project in VR. As displayed in the mockup of the environment, I have created 4 panes namely: surgery, vitals, radiography and chats.

Each of this pane has a custom component designed in blender the actual processing will be handled through code i.e. the materials will be changed on execution as and when they arrive.

The UI buttons are pending as aligning the unity canvas for 2D UI elements is a difficult task and would take some time. Additionally, the UI canvas fails to render in  action based XR rig using the world space camera, anchoring the canvas element to the main camera of the rig solves the issues temporarily.

On a second note, UI based canvas elements could be eliminated for a control system based of a 3D interactor system.

That's all for this week!

Adios!

Low Powered Models for Disease Detection and Classification for Radiology Images

Project Description -

The aim of this project is to create Deep Learning models for detection and classification of radiology images. The models must be compressed such that they can be deployed to low powered devices like ARM devices, Android devices, etc. Compression techniques such as Quantization and Pruning can be used.  

Mentors -

Priyanshu Sinha

Saptarshi Purkayastha

Judy Gichoya

Geeta Priya Padmanabhan

Tech Stack - 

Numpy

Pandas

PyDicom

Tensorflow

Tensorflow-Lite/ Tensorflow-Model-Optimization

Docker

Qemu

Project Link - Click here

Commits - Click here  

Merge Requests - Click here

Why to do this -

There has been a lot of progress in developing Machine Learning models that predict the medical condition of a patient based upon specific inputs relevant to the diagnosis of that condition. However, these models have drawbacks while deployment in real-time on edge devices. Firstly, they have been trained on high-end GPUs that consume a lot of power and have a lot of computational capacity. Edge devices function on limited power and have a considerably low computational limit. Next, these models are extremely large in size, usually a few hundred megabytes. While training, a large amount of space is available. But the same is not reflected on edge devices having low storage capacity. Healthcare professionals do not have high-end machines available for immediate usage of these models. But edge devices, being low-cost, are easily available. To tackle the problem of model deployment, we use model compression techniques that reduce four factors - power consumption, storage usage, computational cost and latency of detection models in the healthcare category.

What have you done -

For the purpose of this project, 2 datasets were used -

RSNA Pneumonia Detection Dataset

Chest-XRay 14 Dataset

The compression techniques used were -

Dynamic Quantization

Float16 Quantization

Int8 Quantization

Model Pruning

Model Pruning + Dynamic Quantization

Model Pruning + Float16 Quantization

Model Pruning + Int8 Quantization

RSNA Pneumonia Detection -

Two models were trained on this dataset - DenseNet201 and InceptionV3. We achieved the following results in the models’ performance with respect to accuracy and size.

Accuracies comparing original and compressed models -

Size comparing original and compressed models -

Accuracies comparing pruned and quantized-pruned models - 

Size comparing pruned and quantized-pruned models - 

Chest XRay14 -

Pretrained CheXNet model was used for this dataset from Bruce Chou's Github repository (link in references). The following results were obtained for this dataset.    

AUROC Score Comparison between orginial and compressed models -

 Model Size Comparison -

How have you done it -

The general pipeline goes like this -

Step 1 - Data Exploration and Cleaning

In this step, we take raw data and explore it. We find out the number of classes, number of data items per class and the general distribution of data points. After deriving these insights, we clean the raw data to get rid of any unnecessary features or data entries. We also restructure tabular data such that it can be fed to the models. This involves steps like creating one-hot encoding of labels, creating extra columns, modifing path variable to redirect to images. In case of images, activities such as augmentation, resizing, shearing, etc are performed. 

Step 2 - Modelling

This is the next step in which we initialize data generators that generate preprocessed images and labels in fixed batches. Data is split into train-val-test subsets. Model architectures are initialized. We have used 3 architectures for this project - DenseNet201, InceptionV3 and CheXNet. We also initialize callbacks, checkpoints and optimizers that will be used during training. 

Step 3 - Training and Model Evaluation

Here, we train the models till we achieve acceptable performance. The model should neither be underfit nor overfit. After training is over, we evaluate the models. We evaluated DenseNet and InceptionV3 trained on the RSNA Pnemonia Detection Dataset based on accuracy. This is because the models directly output the class of the input image. CheXNet trained on Chest-XRay14 dataset was evaluated based on AUROC score because the output was not a fixed class but a class probability score. We also record the size of this original model.

Step 4 - Model Pruning

In model pruning, we trim the unnecessary connections in the neural network. here, I have used Polynomial Decay as the sparsity function. Pruning starts from 50% and goes upto 80% of the total weights in the model. After this, we remove the excess connections and compress the layers of the neural network. This model gets saved in a .h5 format.

Step 5 - Post-training Quantization

After the models are trained, we quantize them. This is done using Tensorflow Lite Converters. There are 3 types of quantizations that we are performing in this project - Dynamic, Float16 and Int8 quantization. We initialize the converter as per our requirement and pass the pre-trained or pruned model to it. The output is a quantized model in the form of a TFLite FlatBuffer. We evaluate the quantized models based on accuracy/AUROC score (as per the original model) and size.

Step 6 - Inference

For performing inference, using normal .h5 models, we use the model.predict() fucntion. While using TFLite models, we initialize interpreters that will set the input and output tensors. We invoke the interpreter on an input image and retrive the output tensor returned by the interpreter. The inference script was run for all models - original, pruned, quantized and hybrid.

Future Scope -

1. Testing these models on actual hardware such as Raspberry Pi and Android phones.

2. Compressing object detection/segmentation/UV based models.

3. Creating a UI to serve these models on the frontend.

References -

1. https://www.tensorflow.org/lite/performance/model_optimization
2. https://youtu.be/3JWRVx1OKQQ
3. https://youtu.be/4iq-d2AmfRU
4. CheXNet - https://github.com/brucechou1983/CheXNet-Keras

• Project Title: Scale-up mHBS Applications.
• Primary mentor: Saptarshi Purkayastha
• Backup mentor: Sherri Bucher
• Student: BHAVESH SHARMA
• Project Link: Project Thread Link

Week 10 Of coding period is completed, I worked on adding Creating sphinx Documentation, Splash Screen, App Icons, Updating program rules, enter server textfield Created Documentation using sphinx Deployed documentation on Gitlab pages Created App Icon Created Splash Screen Updated metadata Created CI script to build documentation Added link to docs in README.md

Link to documentation on GitLab Pages

https://darshpreet2000.gitlab.io/lh-mhbs-eceb/

Splash Screen	App Icon

Link to merge request https://gitlab.com/librehealth/incubating-projects/mhbs/lh-mhbs-eceb/-/merge_requests/25

What progress have you made this week?

1. Prepared Documentation/Readme for App & how to setup
2. Created App Icon, Splash screen, Intro Slides
3. Updating metadata for both programs
4. Finding & fixing bugs
5. Presentation with mentors

What do I plan to do next week?

1. Deploy to F-Droid
2. Finding & Fixing bugs

Have you had any blockers or issues that are impeding your project?

No

Hey Everyone, I am back with my weekly updates about my GSoC journey😀. This amazing journey is about to its end guys, but really I never forget this great experience where I learned a lot of technologies & experienced a new tech side of android development in just a few weeks.

This week I had a meeting with my mentor where I discussed some of the UI-UX related stuff that can make the user experience much better. Hence I have improved the UI of the app, uploaded educational resources over dhis2, find out some very important bugs that need to be resolved in the scale-up version. Let’s discuss them one by one -

Segregated Media Resources into Categories-

• Earlier all types of media files either it is pdf, video, xls file, doc., etc. everyone is present in the single page.
• Since we have more video resources and Pdfs I have made out the filter options to view resources into 4 separate sections All Media, Videos, PDFs and Other files.
• Related Issue: Issue #195 Related MR: MR #199
• Media files are classified based on their mime types.
• Segregation is done only for downloaded media files.

To Classify Documents as :

PDF — pdfMimeTypes = ["application/pdf"]

Videos — videoMimeTypes = ["video/ogg","video/webm","video/.webm","video/.ogv","video/mp4", "video/.mp4","video/.m4v","video/x-flv","application/x-mpegURL","video/MP2T","video/3gpp", "video/quicktime","video/x-msvideo","video/x-ms-wmv"]

Added Metadata of Media Resources & Improved UI-

• Added a button to download and open media files for easy to use purposes.
• Added Thumbnail for video resources & a default pfd icon for pdf files.
• Added video duration in the media card.
• Related MR: MR #200
• Check the new UI & functionalities of the Media Page.

An issue with Cordova plugin Secure storage-

Everything is working fine and in the last few days back I updated my phone and as usual, I keep on changing the code to come up with a better solution. But after some time everything is like 😶.

Suddenly my app gets on crashing on my device and does even not get opened. I was completely clueless why it is happening where the problem is? what I had changed 🤔.

After a long time, I identified the great culprit i.e. cordova-plugin-secure-storage which is begin used in our trainer app to store the user credentials. After doing a google search, I had to find out its cause. Some of the important points regarding this issue and its fix are :

• cordova-plugin-secure-storage has been archived and no longer maintained.
• They are not going to solve the existing issues.
• So I find out some of its alternatives that are cordova-plugin-secure-storage-echo which is a forked version of an earlier one with all of these fixes and being used a lot.
• Therefore I have made the appropriate changes to shift our project from cordova-plugin-secure-storage to cordova-plugin-secure-storage-echo
• Related Issue: Issue #196 Related MR: MR #201
• For More info please check -
• 
  

1. Ionic Native Secure Storage Not Working on Android Q
2. cordova-plugin-secure-storage-echo
3. Crypho/cordova-plugin-secure-storage  — Repo with Issues & archived
4. mibrito707/cordova-plugin-secure-storage-echo — Repo with Fix & active

Uploaded Educational Resources-

We need to upload the real educational resources in the dhis2.

Resources: https://globalhealthmedia.org/videos/

It is having a lot of media files. This week I had uploaded 60+ Media files on the dhis2 and all of these media files are being accessed through the trainer app.

Checked the possibility for opening all types of media files with the app-

It was an idea that it is better if possible to open every media file within our app. Currently, we use the cordova-file-opener2 plugin for this purpose which gives access to the supportive apps to open the file. But the interesting fact to note is that any of the apps do not have access to check the media files and play even they are not visible if we find them on the phone or any supportive app. All the media files are downloaded and saved in users internal storage in a Persistent manner and well encrypted so any of the media files can be open only from the trainer app.

But I checked for the other options -

• We can play video files using the video HTML element.
• The problem is with pdf and another one.
• I checked cordova-plugin-document-viewer. But this is also redirected us to download a specific app from the play store and after that only it can open the pdf using that app. This app uses pdf.js.
• I implemented this and tested but the problem is we need to download an open that app is being used to open pdfs.
• Overall till now also we are using fileopener2 🤷‍♂️.

What is work done this week?

1. Segregated the media files into pdf, video, all media and other options ( Provided filtering options ) 
2. Added Metadata to the Videos UI like thumbnail & time duration.
3. Added Button and smoothen the way to download and view media files.
4. Fixed Issue with Cordova plugin Secure storage-
5. Tested the option to open media files within the app by the cordova-document-viewer plugin & pdf.js.
6. Uploaded Half of the Education Resources over dhis2 for training. Find & Fixed old bugs.

What do I plan to do next week?

1. Work on final submission and improve written blogs.
2. Finding more bugs and fix them.

WEEK 3:

PROGRESS MADE THIS WEEK

Hello everyone. As dicussed in the previous blog, this week was to be utilised for creating the Appointment workflow in the React application.
Surpisingly, the task hardly took 24hrs for completion, henceforth by the end of Tuesday I was done with this week`s work 😄.
Anyways there were a few bugs in test files that were to be fixed, so I utilised my weeks fixing some web-component tests.

here are the screenshots of the workflow:

upcoming days:

• start working on the task of the coming weeks.

next week:

• complete the components for the visit workflow.

see you next week 😄

WEEK 3:

WEEK 2:

Hello folks! Another week done and another batch of code written and merged !! In the coming sections I will be explaining the details of the work done this week.

PROGRESS MADE THIS WEEK

In the previous week, I had implemented the checkin workflow in the react application. According to my plan in the proposal, this week was to be used to create components for the resources that will be used in the Appointment workflow.

the resources are:

• Slot
• Schedule
• Appointment

In the beginning of the week, I started looking into the lit: upon which the components are based. I read the documentation and decided how to create the components so that they can be utilised for different CRUD operations.

I also noticed a few bugs in the components while implementing the checkin workflow, for this I created a MR and fixed them :

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/blob/master/packages/fhir-active-status/fhir-active-status.js#L65 => the value property doesn’t exist on target in this case

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/blob/master/packages/fhir-decease-status/fhir-decease-status.js#L63 : the value property doesn’t exist on target in this case

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/blob/master/packages/fhir-human-address/fhir-human-address.js#L94: the value needs to have a fallback, if text is undefined the component will display undefined

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/blob/master/packages/fhir-human-relation/fhir-human-relation.js#L81: on change doesn’t do expected task

By wednesday I started creating the components. Most of these components have a reference or codeable-concept property datatype, hence creation of components that can be used in different contexts to implement the mentioned datatypes will help a lot.

By the end of wednesday I had created these datatype components :

• reference
• codeable-concept

Once these components were created, the resource components were not difficult to develop. By thursday I had created Slot-Schedule and by friday Appointment.

During this period my mentor noticed a few bugs in the project:

• inconsistent label case use
• unavailable demo files (build issue)

So along with the development of the components, I made a couple of commits for these issues as well.

Template used for component creation:

As a component will be used for CREATE / GET / PUT request , I made it compatible with everything.

For the create part, inside the constructor an initial empty skeleton value was provided. This value property would be overwritten if the component contains the value attribute. Hence this supports both the cases of create and get operations as a user can avoid the value attribute if the intention is to use the component for creating entries and provide a value after fetching a specific endpoint if the intent is to show entries.

Moreover even if a value property is provided and some of the keys are missing, the component will still remain the same as I have done a check for individual property existence and default value allocation as well, this feature will be helpful for the edit operation.

here is the default value property set inside the Appointment component constructor. Similar structure is used in the other 2 components.

1
2
3
4
5
6
7
8
9
10
11
12
13
14

{
      identifier: [{}],
      status: '', 
      serviceCategory: [{coding: [{ system: "", code: "", display: ""}],text: ""}],
      serviceType: [{coding: [{ system: "", code: "", display: ""}],text: ""}], 
      specialty: [{coding: [{ system: "", code: "", display: ""}],text: ""}],
      appointmentType: {coding: [{ system: "", code: "", display: ""}],text: ""},
      slot: [{reference: "", display: "", type: ""}],
      comment: '',
      start: moment().format('YYYY-MM-DDThh:mm:ss[Z]'),
      end: moment().format('YYYY-MM-DDThh:mm:ss[Z]'),
      participant: [{type: [{coding: [{ system: "", code: "", display: ""}],text: ""}], status: '', actor: {reference: "", display: "", type: ""}}]
    }

here are the screenshots of the implemented components:

upcoming days:

• write unit test for the components.

thanks for reading till the end ❤️ , see you next week 😄

WEEK 2:

Hello folks! Another week done and another batch of code written and merged !! In the coming sections I

WEEK 1:

Finally the community bonding priod is over and the coding period has commensed. In this blog I have loads of content and code to cover!

ANALYSIS OF COMMUNITY BONDING PERIOD
In the community bonding period there were mainly 2 objectives we had set :

• SETUP THE REACT APPLICATION WITH CI/CD.
• DESIGN THE APPLICATIONS WITH THE ENDPOINTS.

Most of my community bonding period was spend on the CI/CD part. The major cause of this was
cross-browser, cross-platform testing integration with React. The tool which was planned to be used was saucelabs and unfortunately for me 😕 there was no proper content regarding saucelabs integration with react to be found anywhere, I tried to make things work on my own and nearly succeeded testing on jest with snowpack and react, but snowpack could not resolve the dependencies inside the lh-toolkit-webcomponent monorepo when it was caching the files.

Snowpack serves the application unbundled during development. Each file needs to be -built only once and then is cached forever. When a file changes, Snowpack rebuilds that single file

As mainly we will be covering unit and integration testing, saucelabs will not be a neccessity as suggested by the saucelabs community. So I went on to setup the react app

• testing : using react-testing-library with jest
• deployment : gitlab pages
• some eslint integration and git hooks.

Design part was smooth sailing and offered many insights that I might have not noticed until later, upon the suggestion of my mentor I have also documented the api endpoints. I will drop the file with the designs and the endpoints here : https://geforce6t.github.io/blog/categories/EHR/

PROGRESS MADE IN THE WEEK

the plan for the first coding week is to do the following:

• setup the application
• create home, login pages
• implement the checkin workflow

monday: I started coding 🎉️ , and completed the first part of the plan.

tuesday: I started working on the project layout and accomplished the following :

• organized project into different directories including pages, components, redux store, redux slices etc
• added MUI with theme
• created login and auth pages with support for other workflows in mind.
• created redux auth workflow

wednesday: looked for ways to style the lit-components………, lit components uses shadow dom and the style encapsulation feature make it a little tricky to style custom components from the outside, I made some hacks to do the same using callback refs inside the react component where the lit component resides,but this approach would not be optimal.

Probably some minimal styles could be applied inside the lit components and would do great, anyhow I was able to style the material components using the global material css variables.

One of the things that I need to look into is to add some style to the lit-components as a whole : “margin, border shadow and stuff”, this will not be a big issue as I have a lot of other components to create so mainly the style will be global and can be done anytime.

So here is how the create-patient component has changed visually:

compared to above, this is what we have as default:

thursday: created the general search component.

the general serach component is something that I would like to talk about. The general search component was actually little difficult to implement. For different cases the search params, columns, query parameter may differ so to create a general purpose search component all of this features must be passed as attributes which was done in the implementation,

the functions used in the seach component are

• one for search, based on fetching and storing the result after filtering the result to only include results for the columns using the query.
• another for storing pagination contexts: next page and previous page url,

the most difficult part about the implementation was filtering results to form column enteries, the values that we may want can be nested one/many levels so it is very important to provide a query string to execute this filtering, it is not possible to pass this query string via props as it is inside a map method, so I created a function with switch case for different workflows that we will implement.

finally the components looks like this :

friday: wrote the edit and show patient code to complete the layout of the checkin worklfow 🎉️, but there was one issue that came across while creating the put request, for some of the fhir components the values are not changing even after visually their value is changed, this might be an issue in the react app or the wc repo. I will take a look and solve this issue in the coming days!

UPCOMING DAYS

• add event handlers for the create patient page.
• debug the value not changing issue while making a put request.
• add css for element (not compulsary for now)

thanks for reading till the end ❤️ , see you next week 😄

WEEK 1:

Finally the community bonding priod is over and the coding period has commensed. In this blog I have

PROGRESS MADE THIS WEEK:

Hi, We are almost done with the implemention of the proposal. The additional work is to make improvements and fixes. According to the feedback of the UI/UX designer from the team, the following improvements should be made to the webcomponents:

• Outlined type for all the fields to maintain consistency.
• Appropriate placeholder text.
• Fix spelling mistakes.
• Convert sections to collapsable rows/columns.
• Improve spacing between fields.
• Add asterisks to indicate mandatory fields.

In the coming week these pointers will be implemented along with documentation and other fixes.

DESCRIPTION:

Merge Requests:

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/merge_requests/128

NEXT WEEK’s PLAN:

• Write documentation
• React optimization
• find and fix bugs.

see you next week 😄.

Video Server OpenCV pipeline for both segmentation and object detection was completed. The Mock HMD VR designed using Unity XR refers to a remote location on the server for getting video output.

Apart from the server design. documentation was completed for WebXR approach. For UnityXR based method few sections of documentation is still pending. For loading model for inference in OpenCV pipeline Frozen_Graph approach is used.

The video server generates close to 5200 images from a Video of length 3 minutes and 29 seconds. The number of images generated depends on the frame rate used. The frame rate has to be optimally adjusted according to the system configuration.

Similarly, the video output of the program depends on the frame rate selected. The frame rate changes video length.

Sample Model Inference :

That's all for today!!!

Hope you had a great week

Tot ziens

Hey Everyone, this week I have fixed the tracker app pipeline that failed last week, exported the tracker and trainer specific metadata, tested them by importing them to the play servers and a lot more. Let’s discuss this in detail.

Fixed Tracker App Gitlab Pipeline

Last week when I submitted an MR that passed the pipeline and was merged by my mentor into the main repo of the tracker app but it gets failed unexpectedly.

The interesting fact to note is that it ran successfully in the commit with which I had submitted but the merge commit that my mentor made to merge it which doesn’t contain any new changes, and is only having the commits that I made failed the pipeline. I didn't get any reason for it.😌

But in the end, I made it find out the cause of it and fix it😀. I think it is due to the NDK version. I suggest if you are building it locally as well please try building after installing NDK Version 21.0.6113669

Related MR: MR-15

Exported Tracker App Metadata

I had exported tracker specific metadata so that one can import them in his own dhis2 instance and can get benefitted from the use of the mhbs — tracker app.

I have exported 2 mHBS Tracker App-specific metadata-

• Exported HBB Survey Metadata
• Exported HBB Program Metadata

To make both of the metadata importable I had deleted the old references from the metadata like “user”, “lastUpdatedBy”, “organisationUnit ”keys from the corresponding JSON. There are more than 4k+ such instances in the HBB Program metadata as it was heavily used by the bmgfdev instance users. 

Now the metadata can be imported into the new dhis2 instance easily.

Corresponding Issue : Issue #22

Corresponding MR : MR-16

Discussion Thread : Link

Import Testing of Metadata

I had tested the import of the metadata over Dhis-2 Play Severs with all data elements and attributes. There is 2 issue that my mentor suggested me to test with the export and import related to HBB Survey are also tested and can be closed as a FIX.

Issue #12 

Issue #11

What progress have you made this week?

• Fixed the tracker app pipeline that failed last week
• Exported HBB Survey Metadata & HBB Program Metadata
• Tested Importing of testing Metadata
• Can be Closed — Issue #12 Issue #11
• Issue #22 — Export mHBS tracker and trainer metadata from DHIS2
• MRs : MR-16 & MR-15

What do I plan to do next week?

• Discuss with the mentor the features and issues to work on.
• Fix more issues, check for current needs.

Week 9 Of coding period is completed, I worked on adding notification count, About app, Share app feature, Load more notifications functionality

Notification count

I created notifications count in bottom app bar, It displays the count of new notifications in app, After fetching data from dhis2, app check if it has the message id , if it doesn’t have the id then count is incremented.

When user visits the profile page then count becomes 0 as the messages are read by the user now.

About app

It displays about the project with a button to visit the project, This screen will be useful for promoting LibreHealth organization to users who are using this application.

Load more notifications

When user click on load more button then it fetches the next 5 notifications, It calls api with next page number & save the data to hive storage.

Link to merge request https://gitlab.com/librehealth/incubating-projects/mhbs/lh-mhbs-eceb/-/merge_requests/24

What progress have you made this week?

1. New notifications count viewing on bottom bar
2. About App, Share App, Report bug menus in drawer
3. Added Load more feature on notifications

What do I plan to do next week?

1. Prepare Documentation/Readme for App & how to setup
2. Creating screen to add server url
3. Individual/facility login
4. Creating App Icon, Splash screen, Intro Slides
5. fetching program rules for classification function
6. Updating metadata for both programs
7. Finding & fixing bugs
8. Deploy to F-Droid

Have you had any blockers or issues that are impeding your project?

No

PROGRESS MADE THIS WEEK:

Hi, As discussed in the last week’s blog, Week 8 was to be used for the following tasks:

• Completing the visit-provider workflow.
• Completing the e-prescription workflow.
• fixing build fail.

DESCRIPTION:

The above mentioned points were covered this week.

visit-provider workflow : this workflow was very much similar to the visit nurse workflow that was implemented last week.

• The provider first finds the patient.
• The provider selects the encounter under which the meeting will proceed.
• The provider then gets directed to the dashboard where he/she can see the patient’s already existing observations and can create an observation report.
• the provider can also create an order request.

Out of these the first 2 points were already implemented in the visit-nurse workflow.

The rendering of different screens is done using redux, each click when required changes a redux state that renders a different screen.

here is the reducer for this workflow:

The editActivePage reducer is for changing the screen. The observation results shown require the Patient Id hence the Id is stored after the provider selects the patient.

All of these screens are a part of a custom High order Navbar component that renders different tabs based on state values.

here is the navbar component :

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181

import React, { useState, useRef } from 'react'
import PropTypes from 'prop-types'
import { useStyles } from './style'
import {
  AppBar,
  Tabs,
  Tab,
  Typography,
  Box,
  CssBaseline,
  Container,
  Button,
  Dialog,
  DialogActions,
  DialogContent,
  DialogTitle
} from '@material-ui/core'
import ExittoApp from '@material-ui/icons/ExitToApp'
import logo from '../../assets/images/logosm.png'
import {
  logout
} from '../../reducers/auth/authSlice'
import '@lh-toolkit/fhir-encounter/fhir-encounter.js'
import { useDispatch } from 'react-redux'
import { useHistory } from 'react-router-dom'
import { toast } from 'react-toastify'
import axios from 'axios'
import { config } from '../../config'
import 'react-toastify/dist/ReactToastify.css'

function TabPanel (props) {
  const { children, value, index, ...other } = props

  return (
    <div
      role="tabpanel"
      hidden={value !== index}
      id={`simple-tabpanel-${index}`}
      aria-labelledby={`simple-tab-${index}`}
      {...other}
    >
      {value === index && (
        <Box p={4}>
          {children}
        </Box>
      )}
    </div>
  )
}

TabPanel.propTypes = {
  children: PropTypes.node,
  index: PropTypes.any.isRequired,
  value: PropTypes.any.isRequired
}

export const Navbar = ({ labels, component, encounterID, includeEncounter }) => {
  const classes = useStyles()
  const [value, setValue] = useState(0)
  const dispatch = useDispatch()
  const history = useHistory()
  const [open, setOpen] = useState(false)
  const [disablepost, setDisablePost] = useState(false)
  const reference = useRef()

  const handleClickOpen = async () => {
    setOpen(true)
    try {
      const res = await axios.get(`${config.basename}Encounter/${encounterID}`)
      if (res.status === 200) {
        reference.current.value = res.data
      }
    } catch (error) {
      console.log(error)
    }
  }

  const handleClose = async () => {
    try {
      setDisablePost(true)
      reference.current.value.resourceType = 'Encounter'
      reference.current.value.id = encounterID
      const res = await axios.put(`${config.basename}Encounter/${encounterID}`, reference.current.value)

      if (res.status === 200) {
        toast.dark('EDITED SUCCESSFULLY !')
        setOpen(false)
      }
    } catch (error) {
      // eslint-disable-next-line node/handle-callback-err
      error.response.data.issue.forEach(err => {
        toast.error(err.diagnostics)
      })
    } finally {
      setDisablePost(false)
    }
  }

  const handleChange = (event, newValue) => {
    setValue(newValue)
  }

  return (
    <Container component="main" maxWidth="lg">
      <CssBaseline />
      <div className={classes.paper}>

        <div>
          <img src={logo} alt='logo'/>
        </div>

        {includeEncounter
          ? (
          <div className={classes.exit}>
          <Button
            aria-label="submit"
            type="submit"
            variant="text"
            color="secondary"
            className={classes.submit}
            onClick={handleClickOpen}
          >
            EDIT ENCOUNTER
          </Button>
          </div>
            )
          : ''
        }

        <div className={classes.exit}>
          <button onClick={() => {
            dispatch(logout())
            toast.dark('LOGGED OUT !')
            history.push('/')
          }} className={classes.logoutbtn}
          >
            <ExittoApp />
          </button>
        </div>

        <AppBar position="static" color="default" className={classes.navbar}>
          <Tabs value={value} onChange={handleChange} aria-label="checkIn tabs">
            {labels.map((item) => <Tab key={item} label={<Typography variant="subtitle1">{item}</Typography>} />)}
          </Tabs>
        </AppBar>

        {component.map((Component, index) => {
          return <TabPanel key={index} value={value} index={index}><Component key={index} /></TabPanel>
        })}
      </div>
      {
        includeEncounter
          ? (
          <Dialog open={open} onClose={handleClose}>
        <DialogTitle id="form-dialog-title">EDIT ENCOUNTER</DialogTitle>
        <DialogContent>
            <fhir-encounter ref={reference} />
        </DialogContent>
        <DialogActions>
          <Button onClick={handleClose} disabled={disablepost} className={classes.submit} color="secondary">
            SUBMIT
          </Button>
        </DialogActions>
      </Dialog>
            )
          : ''
      }

    </Container>
  )
}

Navbar.propTypes = {
  labels: PropTypes.array,
  component: PropTypes.array,
  encounterID: PropTypes.string,
  includeEncounter: PropTypes.bool

The e-prescription workflow is almost similar to the visit-provider workflow, other than 1/2 screens.

• The prescriber first finds the patient.
• The prescriber selects the encounter under which the meeting will proceed.
• The prescriber then gets directed to the dashboard where he/she can see the patient’s allergies and medication and can create a medication request.

Merge Requests:

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/merge_requests/127
• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/-/merge_requests/126

Implementation video:

• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/uploads/22cb8a9f4292fb2ac8b127f20a983fdf/e-prescription.mp4
• https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/uploads/0dbb1220abdbc1a427c2a9c765f8f84a/visit-provider.mp4

NEXT WEEK’s PLAN:

• Write documentation
• Add extra test files.
• React optimization
• find and fix bugs.

see you next week 😄.

Prototype for immersive was designed in Unity using inbuilt XR plugins. For HMD emulation MockHMD was used. The material was rendered on the plane. The material was linked to 2d texture. The VideoPlayer asset in Unity helped in converting the frames to textures. The camera was bounded to the camera space. There were no performance issues while development and testing. The asset used was present locally, asset fetched using URL can also be used.

Segmentation Model from previous week, crashed despite efforts. So, I created a UNet from scratch. Basic model performs well for sample space. Hyper-parameter optimisation is pending. For 300 epochs a dice coefficient of 0.84 was achieved. This UNet model works with openCV. For developing the OpenCV segmentation pipeline only the model inference and drawing portion has to change. Inference takes close to 1~2 seconds for 512x512x3 image.

That's all for today!!!

Hope you had a great week

Farvel

Hello Everyone, I am elated to share that project is going good and mostly all the issues are being sorted in the GitLab. This week I had tested the workflow of the apps, fixed few bugs and updated Readme and added some demonstration videos to help in the use of mhbs apps.😀

Tested or Fixed issues 

• Issue #14 — View data completion form

I have verified that there is a functionality in the mhbs Tracker app to mark the existing enrollment as complete and even there is also an option to re-open the completed enrollments as well.

To understand the working please check this.

• Issue #13- Create “seconds” data element for OSCE B

I have checked the second data element presence and working in OSCE B form. I had tested few other things as well. The data elements that I had tested and present in OSCE form are :

1. Start time -> (TIME type. (hh:mm))
2. Ventilates with bag and mask within The Golden Minute (at ___seconds) -> (fill in the blank)
3. Time of effective ventilation (chest moving gently at __seconds)-> (fill in the blank)
4. Ventilates at 40 breaths/minute (30–50 acceptable) -> (yes only)

Link of the OSCE B Form: 

Link Check_OSCE_B_form_of_already_existing_event

Create_new_Event_in_OSCE_B

• Issue #10- Tested all the features that need to be present in the tracker app

All the features of Module_1_mHBS_General_User_Guide_JANUARY_30_2019_FINAL.pdf are being tested and present in the new tracker app.

But the new UI is entirely different and features may not be located in the place where they are shown in this file. I am thinking of making a new user guide for the tracker app and will discuss that regards with the mentor.

• Issue #11- Need to change frequency of tracker forms and HBB survey.

I had checked the frequency of the existing HBB Survey and it is as per our need but mentors might want me to check the same for exported metadata. Till now I had not completed the work of exporting and importing metadata when it gets completed it will be fixed at that time.

• Issue #12- Need to modify HBB infrastructure survey.

Similar to the above issue, I had checked its functionality with the existing HBB Survey and it is fine. We should also need to check it with exported metadata as well and will do that when exporting part gets completed.

Modify Syncing Logic for Media files.

There is a slight modification being made in the regular syncing of media files as per discussion with mentors. 

Previous Approach: Whenever the user clicks on any media file and if he is connected to the internet we will re-download the file and save replace it with its old locally saved copy. It was done to make sure that we always have the latest and correct file.

Problem With the Previous Approach: There would not be regular updates on the media file but whenever the user checks that media file it will get downloaded in the background and consumes a lot of the user’s data.

New Approach: As per discussion with the mentor we had decided that we will never update the existing media file if we need to update this file then we will remove it and upload a new copy with the same name in this way we never need to download the file again and again. However, you can change the name of an existing file, add a new file, remove old files but not updated the media content of the existing file.

As per the above discussion, I had made changes to the trainer app in order to stop downloading media content again and again. There are few bugs that I encountered during testing in the trainer app and has been fixed.

Committed : b8f6403d

Related Issue: Issue #14

Updated Readme file regarding Media Resources

I had updated the readme of the trainer app regarding all these logic and added resources like 

Demonstration_of_Uploading_resources_on_dhis2

Demonstartion_of_accessing_resources_on_trainer_app_through_tracker_app

Committed : 8483042a 

Related MR: Link

Works Done this week.

1. Tested or Fixed issues -

• Issue #14 — View data completion form
• Issue #13- Create “seconds” data element for OSCE B
• Issue #10- Tested all the features of Module_1_mHBS_General_User_Guide_JANUARY_30_2019_FINAL.pdf in the new tracker app.
• Issue #11- Need to change frequency of tracker forms and HBB survey.
• Issue #12- Need to modify HBB infrastructure survey.

2. Fixed Issue #14 — Modify Syncing Logic for Media files.

3. Updated Readme — Added Documentation demonstrating to access media resources through trainer app and upload resources through dhis2.

    Committed 8483042a 

4. Related MR: Link

5. Small clips to depict the use of the media page.

Demonstration_of_Uploading_resources_on_dhis2

Demonstartion_of_accessing_resources_on_trainer_app_through_tracker_app

Work to be done Next Week

1. Fixing tracker app GitLab CI failed last week.
2. Export metadata of tracker and trainer app and add to the readme.
3. Some testing in HBB Survey after exporting metadata.
4. Issue #14 — Create BMGF Indicators
5. Issue #44 — Role-based restricted access for viewing tracked entities.
6. Discuss with mentors about the complete app for further scale-up.

Have you had any blockers or issues that are impeding your project?

Export and Import of metadata and GitLab CI failed last week.

I want to discuss them with my mentor and want to get fix them.

Week 8 Of coding period is completed, I worked on adding on call doctors schedule showing screen, I created stage 5 assessments capture feature

Added On-Call Doctors Schedule Showing feature

Tasks

• shows list of doctor in program
• shows name, date, start time , end time of on call schedule
• shows profile image in circle
• uses cache image so that user can view image & data offline

Created Stage 5 Discharge with floating button

• Create Discharge assessments UI
• Created discharge floating action button to appear only when child is normal & 24 hours have passed
• Create model class for discharge
• Create stage 5 discharge event
• added methods to validate , register stage 5
• added api request to mark enrollment as COMPLETED
• added method to queue api request when user is offline
• Created unit test for stage 5
• Created stage 4 variable schedule to store time

Stage 5 Unit test cases

• Throws exception if assessments are incomplete
• Marks phase-5 as complete if everything is correctly filled
• Marks child as discharged
• yields AssessmentsAdded on successful adding of phase 5
• yields AssessmentsError when assessments are incomplete while adding of phase 5 assessments

What progress have you made this week?

1. Fetching & displaying doctor’s schedule in App with unit tests
2. Created Stage 5 (Discharge) with button to discharge based on conditions
3. Created unit test for stage 5
4. Made stage-4 repeatable in app after 180 minutes

What do I plan to do next week?

1. Making messaging/Notifications viewing functionality in App
2. Adding share App report bug menus in drawer
3. Preparing Documentation/Readme of Project

Have you had any blockers or issues that are impeding your project?

• I have completed assessments capture functionality of app, I will request feedback from mentors & will work on the same.
• I need to discuss individual/facility login feature

PROGRESS MADE THIS WEEK:

Hi, As discussed in the last week’s blog , Week 7 was to be used for the following tasks:

• Complete the visit-nurse workflow (the parts which are not implemented yet.)
• Complete the visit-provider workflow.
• Start working on the e-prescription workflow.

This week I was a little busy with some academic work, hence all of the above pointers were not possible, Hence the last 2 pointers will be shifted to the upcoming week. (WEEK 8)

DESCRIPTION:

This week I have mainly worked on completing the visit nurse workflow. As the vitals signs and medication statement components were merged, the react-ehr application was updated to include the screens which use these components in the visit-nurse workflow.

Another interesting thing is that most of these screens are having almost a similar structure, hence there is a possibility that a single component can be created as a layout component for different screens.

NEXT WEEK’s PLAN:

• Complete the visit-provider workflow.
• Start working on the e-prescription workflow.
• fix build fail.

The dosage component seems to cause a build failure, which will be fixed in the coming week.

reiterating the pointers from last week’s blog for the implementation points of the visit-provider workflow:

• First screen contains a find-patient component to select a patient.
• After the patient is selected a redux action is triggered which updates redux state and stores the selected patient Id.
• Another redux action updates the active screen and changes it to that of patient encounters.
• The patient encounter screens contains all the patient encounters which are fetched using the stored patient Id.
• On clicking an encounter a redux action updates active page to dashboard.
• the dashboard page contains 2 tab: physical exams, order.
• the physical exams tab shows the observations of the patient and contains a component to create any observation.
• the orders tab will be used to create orders using service request component.
• the user can log out and select a different patient to repeat the above points.

the pointers for the implementation points of the e-prescription workflow:

• First screen contains a find-patient component to select a patient.
• After the patient is selected a redux action is triggered which updates redux state and stores the selected patient Id.
• Another redux action updates the active screen and changes it to that of patient encounters.
• The patient encounter screens contains all the patient encounters which are fetched using the stored patient Id.
• On clicking an encounter a redux action updates active page to dashboard.
• the dashboard page contains a single screen with a search component for allergies and medications of the patient.
• the screen also contains a create medication request component.

see you next week 😄.

Tested 3–4 Issues

I had go through almost all the issues of the tracker and trainer app. I had tested issues 4 issues marked with testing label and prepared their solution. I will be discussing my approaches for these issues with my mentor and fixed them in next week.

Fixed mhbs Logo Issue Issue #4

In the tracker app some of the screens have the dhis-2 logo. It is now fixed to mhbs logo.

Fixed Issue #4 

Related MR : Mhbs logo fixed

Updated Trainer App Readme.md

Since we had added media page, offline support, app usage tracker in trainer app we want a small documentation details to be added to the readme for further reference. I had done it -

Committed here ( 718f8756 ).

Fixed Bug in GitLab CI

There is a bug in GitLab CI code of trainer app. Gradle is building the apk correctly but it is not assembling the artifact correctly. It is my mistake what I had made earlier. I had corrected it.

Committed f01c6cf9.

Work Done this Week

• Fixed logo Issue. 
• Updated Trainer App Readme.md.
• Fixed Bug in GitLab CI.
• Tested 3–4 Issues & prepared their solution. After discussing with mentor need to implement them.

Planning of next Week

• Need to modify HBB infrastructure survey
• Need to change frequency of tracker forms and HBB survey
• Fixing more issues which I had tested this week and prepared solution to fix them.

The video object detection pipeline is designed using openCV. The videos are loaded and a frame rate of 25FPS is assumed. Using the frames the tf.session is initiated and TF Object detection API model is run. The resulting bounding boxes are drawn using openCV.

Resulting images with bounding boxes are written to video with 1FPS using openCV videowriter. After that the generated video can be served using data or blob url depending upon the video so generated.

The segmentation model is not working properly for most of the cases. The obvious solution for it would be to retrain the entire model. The segmentation model has to be implemented in the same way as the Object detection using openCV.

At present the entire video generation pipeline takes less than 40 seconds. This excludes the uploading time and downloading time. The server on which it was tested has a RTX 3060 GPU peak GPU memory utilisation reached upto 11GB

That's all for today!!!

Hope you had a great week

Annyeong

Week 7 Of coding period is completed, I worked on adding on call doctors viewing feature, I completed notifcation showing functionality on risk assessments and monitoring alerts

Added On-Call Doctors Viewing Functionality in App with Unit tests

• Fetching events from on call doctors program
• displaying on call doctors
• displaying images with cache
• saving the list to local storage

Unit tests

• yield OnCallDoctorLoaded on successful fetching of data
• yields OnCallDoctorLoaded on unsuccessful fetching of data by getting from hive
• returns response with 200 code on successful fetching of on call doctors
• throws fetch data exception on no internet connection

GIF Showing on Call Doctors slider

Completed Notification screen (Risk Assessment, Monitoring Alerts Tab) with Unit tests

• Created notification screen bloc
• Created repository with methods to get risk assessments and monitoring alerts
• on tap of the list item the child assessments are opened

Unit tests

• Flutter Notifcation BloC test
• Test Cases
• yield NotificationLoaded on Fetch Nofitication Of Babies
• Flutter Notifcation Repository test
• Test Cases
• Risk Assessments Test cases
• status changed normal to danger
• status changed problem to danger
• status changed normal to problem
• status changed danger to problem
• Monitoring Alerts Test cases
• stage-4 assessments next

Unit tests Phase 4

AssessmentsRepository Test Cases

• Throws exception if assesments are done before 180 mins from birth
• Throws exception if assessments are incomplete
• Marks phase-4 as complete if everything is correctly filled
• Adds stage-4 assessments if list is empty

AssessmentsBloc Test Cases

• yields AssessmentsAdded on successful adding of phase 4
• yields AssessmentsError when assessments are incomplete while adding of phase 4 assessments

Phase 3 Assessments Unit testing

Test Cases

• Throws exception if Stage3 Normal assessments are incomplete
• Throws exception if Stage3 Problem assessments are incomplete
• Throws exception if Stage3 Danger assessments are incomplete
• Marks Stage3 Normal as complete if everything is correctly filled
• Marks Stage3 Problem as complete if everything is correctly filled
• Marks Stage3 Danger as complete if everything is correctly filled
• Adds Stage3 Normal assessments if list size is 2 and classification is normal
• Adds Stage3 Problem assessments if list size is 2 and classification is problem
• Adds Stage3 Danger assessments if list size is 2 and classification is danger

Bloc Test

• yields AssessmentsAdded on successful adding of phase 3
• yields AssessmentsError when assessments are incomplete while adding of phase 3 assessments

What progress have you made this week?

• Added On-Call Doctors Viewing Functionality in App
• Completed Notification screen (Risk Assessment, Monitoring Alerts Tab)
• Phase - 3,4 Unit testing

What do I plan to do next week?

1. Fetching & displaying doctor’s schedule in App with unit tests
2. Making messaging/Notifications viewing functionality in App with unit tests
3. Creating share App report bug features
4. Creating Stage 5 (Discharge) Stage with unit tests

PROGRESS MADE THIS WEEK:

Hi, According to the proposal timeline, week 6 was to be based on:

• finishing pending work.
• fixing bugs.
• starting week7 tasks.

Fortunately, there was no pending work, hence the focus was mainly on the remaining points.

The tasks of week7 are to:

• implement the visit-nurse and visit-provider workflow in the react application.

As most of the components required for implementing the visit-nurse workflow are already present in the web-component library, I started working on it.

DETAILED DESCRIPTION:

The visit-nurse workflow consists of many screens and multiple components.

• First screen contains a find-patient component to select a patient.
• After the patient is selected a redux action is triggered which updates redux state and stores the selected patient Id.
• Another redux action updates the active screen and changes it to that of patient encounters, it also updates redux state and stores the current encounter Id.
• The patient encounter screens contains all the patient appointments, encounters and a component to create an encounter for the undergoing visit.
• On creating the encounter a redux action updates active page to dashboard.
• the dashboard page contains 4 tab: home, vitals, allergies, medication.
• the home tab is the default active tab and contains already exisiting vitals, allergies and medications of the patient.
• the other tabs are used to create specific resource entry.
• the user can log out and select a different patient to repeat the above points.

Some of the required components are not merged in the main repo yet, hence respective tab is unimplemented.

Each worklow in the application is a different page with a unique url. Hence the use of redux in immense in the project.
Each workflow also has a dedicated reducer which handles all the operations related to a specific workflow without editing any other state.

Inside the project’s src directory, the workflow directory contains all the code distributed into separate directories for different workflows. This makes it very simple to create / edit or even delete a workflow without affecting other parts of the application!

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

import React from 'react'
import { Navbar } from '../../components/navbar'
import { Home } from '../../pages/home'
import { activePageName } from '../../reducers/visit/visitNurseSlice'
import { CreateEncounter } from './pages/createEncounter'
import { SelectPatient } from './pages/selectPatient'
import { Dashboard } from './pages/dashboard'
import { useSelector } from 'react-redux'

export const VisitNurse = () => {
  const activePage = useSelector(activePageName)
  let components
  switch (activePage) {
    case 'SelectPatient':
      components = [Home, SelectPatient]
      break
    case 'CreateEncounter':
      components = [Home, CreateEncounter]
      break
    case 'Dashboard':
      components = [Home, Dashboard]
      break
  }
  return (
    <Navbar labels={['HOME', 'VISIT']} component={components} />
  )
}

This is the main component that is exported. Based on different redux state values different screens are displayed.

All the implemented screens and the workflow can be seen here:

https://gitlab.com/librehealth/toolkit/lh-toolkit-webcomponents/uploads/a9e04307d1199767e3b2c400de785370/screen-capture__2_.webm

NEXT WEEK’s PLAN:

• Complete the visit-nurse workflow (the parts which are not implemented yet.)
• Complete the visit-provider workflow.
• Start working on the e-prescription workflow.

For the visit-nurse workflow, the medication statement and the vitals tabs are left to be implemented, these must be completed within the next week.

There are also some additions to be made in the webcomponents : some fields like encounter and subject (patient) are required to be added to the allergy and medication statement components.

The visit-provider workflow has similarities to the visit-nurse workflow, some of the screen are similar for both, hence the provider workflow should be completed faster than the nurse workflow.

these are the implementation points for the visit-provider workflow:

• First screen contains a find-patient component to select a patient.
• After the patient is selected a redux action is triggered which updates redux state and stores the selected patient Id.
• Another redux action updates the active screen and changes it to that of patient encounters.
• The patient encounter screens contains all the patient encounters which are fetched using the stored patient Id.
• On clicking an encounter a redux action updates active page to dashboard.
• the dashboard page contains 2 tab: physical exams, order.
• the physical exams tab shows the observations of the patient and contains a component to create any observation.
• the orders tab will be used to create orders using service request component.
• the user can log out and select a different patient to repeat the above points.

see you next week 😄.

WEEK5

PROGRESS MADE THIS WEEK:

Hi! We have a small change in the implementation plan. As per the proposal this week => week 5 was meant for implementing the visit workflow in the react / EHR application. The implementation required some new components, which were created last week as per the plan, but the Merge Requests are not merged yet. As there can be multiple changes and modifications required after the MRs are reviewed, It would be better to wait till the components are merged after modifications if required.

Therefore this week was based on doing the tasks of the 7th Week as Week6 is planned for improvements and bug fixing.

Primarily this week was spent creating the last set of components that will be required for the workflows planned.

The following components were created -

• MedicationRequest
• dosage

These components will be used for e-prescription workflow, although the dosage backbone element can be used for other Medication workflow based resources as well.

The pattern used for creating these components is same as the one used for other components.

Here is the screenshot:

next week:

• find and fix bugs in the react app.
• start implementing visit workflows in the react application.

see you next week 😄.

WEEK5

The main focus for this week was correcting the errors in the bounding box drawn. The bounding boxes differed from the ground truth. That was because of issue regarding the canvas context drawing setting and not related to the actual model.

After the correcting the drawing steps in the canvas context. The bounding box are drawn somewhat similar to the ground truth.

Second goal for this week was development of a remote server based object detection video api. I successfully mad it work for single image but beyond it the server started to hang-up.

The video encoding and decoding process are particularly stressful for the CPU. Throughout the development time the CPU was under 96~97% utilisation running at maximum clock speed. This was the case for a single instance, but if multiple instances are launched the server crashes instantaneously. For running the frozen inference model GPU was used. Under single gpu setup the gpu was under 100% memory utilisation.

Apart from this network latency spikes up during the encoding and decoding process.

That's all for today!!!!

Hope you had a great week

Adios

Settings drawer to toggle settings with unit test

• Created end drawer to toggle settings
• User can toggle notifications, notifications sound, dark mode
• The settings done by user are saved instantly in storage and changes are made according to it
• Wrote unit test for settings bloc with test cases 1. yields SettingsState on toggle of dark mode 2. yields SettingsState on change of notifications sound 3. yields SettingsState on toggle of notifications

Stage 3 Assessments Capture functionality

• Stage 3 has 3 types (normal, problem, danger)
• If a baby is classified as any type then that type of stage 3 assessments get added in app

Video GIF showing stage - 2 with classification functionality

• background turns red on danger
• background turns yellow on problem
• background turns green on normal

Stage 2 Unit testing

Test Cases for stage-2

• Throws exception if assesments are done before 60 mins from birth
• Throws exception if wardname is empty
• Throws exception if assessments are incomplete
• Marks phase-2 as complete if everything is correctly filled
• Adds stage-2 assessments if list is empty

Test Cases for Classification Repository

• if temperature < 95.5 or > 99.5 then classifies as danger
• if severe jaundice or fast breathing then classifies as danger
• if weight < 2000 & poor feeding then classifies as problem
• if everything is normal then classifies as normal

Showing summary of 24 hours in home screen

To show summary of 24 hours, We need to use lastUpdatedDuration=1d (1 day) as parameter in API to fetch all the events happened within 24 hours.

This has 3 types

• Admitted Count - Check Status = ACTIVE & Enrollment Date yesterday
• Discharged - Check Status = COMPLETE
• High Risk - For each TEI (Baby) this needs to be counted by parameter classification

What progress have you made this week?

• Phase - 3, 4 assessments capturing functionality
• Showing summary in Home Screen
• Developing settings screen of app
• Unit Testing phase - 2

What do I plan to do next week?

• Adding On-Call Doctors Viewing Functionality in App
• Completing Notification screen (Risk Assessment, Monitoring Alerts Tab)
• Phase - 3,4 Unit testing

Hey Everyone, I am back with my weekly blog. Let’s have a look at work done in the 6th week of this amazing journey.

Reimplemented and Completed App Usage Tracking System

Guys actually as stated in the last blog I had completed the app usage tracking system but on discussing with the mentor We come up with a better approach and reimplemented the completed system. There are not very huge changes we just did some changes in the dhis2 storage part. Let’s have a look at how we are tracking the trainer app usage now -

• App usage is now stored in the user’s data store instead of the tracker program.
• For each page, a unique id is assigned with which the page usage is tracked in datastore and in local DB as well.
• There is a local DB support that will store the page usage and when the page usage exceeds the threshold limit for each usage then data will be sent to the dhis2 in the user’s datastore.
• When local app usage storage get synced with the dhis2 user’s data store then local DB will get resettled to store new data.

Related MR: Added System to track app usage (!196) · Merge requests

Resolved Issues: #36 #120

Passed Pipeline:  #338127000 

Modified mhbs-tracker app to send login credentials

Since we want the user to enter credentials only once for both the apps and trainer app to be launched from the tracker app. Therefore the idea is, tracker app will share the login credentials with the trainer app on each launch. I had saved the login credentials of the user in secured Shared Prefs and on launching intent of trainer app passed these credentials so that trainer app to process further.

Related MR: Passed user credentials to trainer app for login & Fixed Some Warnings · Merge requests

Resolved Issues: #16

Passed Pipeline:  #338443846

Removed Fabric usages

The fabric plugin is depreciated and we need to remove its instances from the app.

Committed here.

Other Works

• Added commits to the previously sent MR as asked by the mentor to make them able to merge. As for testing purposes, I had used the hardcoded credentials, Now I had removed them and fetched the password from secure storage.
• I am not clear with the login procedure of the trainer app. I had studied the codebase and understand it well and made changes in the trainer app accordingly to make the user able to get logged in to the trainer app from the tracker app.

Work Done this Week

1. Implemented New Idea suggested by the mentor and Completed the system to get trainer app usage over dhis2. 
2. Removed Fabric usages from the app as it is depreciated.
3. Modified mhbs-tracker app to send login credentials to trainer app for login.
4. Added commits to the last MR’s to make them able to merge as suggested by the mentor.
5. Completely understand the secure storage and login flow of the trainer app.

Planning of next Week

1. Completed all pending work so can focus on new issues. Will be solving more bugs.
2. Working on “Need to create a new program stage in dhis2” issue of tracker app.
3. Working on #4 issue.

I am not clear with these issue so will try to discuss these issue in the weekend so that can start work from Monday.

Important Links

1. Added System to track app usage (!196) · Merge requests
2. Resolved Issues: #36 #120
3. Passed Pipeline:  #338127000 
4. Passed user credentials to trainer app for login & Fixed Some Warnings · Merge requests 
5. Resolved Issues: #16
6. Passed Pipeline:  #338443846

Hello Everyone, sorry for getting delayed in writing this blog. Last week's work is as follows: 

Worked on App Usage Tracking System of Trainer App 

To track the app usage I had followed these steps-

1. Created “mHBS Trainer Usage Tracker” Tracker Program

• Created Tracked Entity Attributes to store [ total page visits, total time spent on a page, page visits per user, time spent on a page by per user].
• Created “Page Usage Tracker” Tracked Entity with the above attributes.
• Created a tracker program i.e. “mHBS Trainer Usage Tracker” to track the tei created above (Page Usage Tracker).
• There are more than 40 pages in Trainer app. Enrolled tei for each page in “mHBS Trainer Usage Tracker” program.

Now we can send data for each and every page to dhis2 in the Tracker program.

2. Created Local SQL- DB table to store app usage

• Users cannot remain online every time but we want the full usage of our app to track therefore we need some local storage to store app usage in offline mode and update this data to dhis2 whenever the user comes online.
• Therefore created “app_usage” SQL table in the trainer app with the — [page_id (stores tei id of the page with which it is created at dhis2), page_name, page_visits, time spent on page ]
• Whenever a user visits any of the pages ( get to know from routes ) then we update its visits in local DB. 
• And similar to time spent on any page, it will be calculated with the help of “beforePageIn” & “beforePageOut” Events with simple subtraction of time in invoking this event.
• Then update this data to dhis2

3. Idea of Threshold Limit for app usage

• I thought that instead of hitting the API continuously to update dhis2 with current usage we can make a threshold limit like 50-page visits means whenever any page visits crossed the number 50 only in that cases we will hit the API to update dhis2 and in other cases, we just update the usage in local DB.
• This will help us to save a lot of users data.

All of the above-discussed flows are integrated and set up. But recently we come up with a better approach suggested by the mentor, so I will try to implement that one and send the MR instead of this approach.

Works Done in the Fourth Week

• Created “Page Usage Tracker” Tracked Entity & “mHBS Trainer Usage Tracker” Tracker Program
• There are approx 40 pages in the trainer app, made tei instance for every page in “mHBS Trainer Usage Tracker” program to keep its usage track.
• Stored page visits and time spent on each page by the user in local SQL-DB as well as updated this to dhis2 program. 

What do I plan to do next week?

• Complete Pending Tasks, reimplement this solution with a better approach suggested by the mentor.
• Try to fix some more bugs.

Have you had any blockers or issues that are impeding your project?

• Not yet

The main target for this week was POC development. I was successful able to develop a POC with help of HTML5 Canvas and WebXR. The total inference time is about 25 seconds on a Snapdragon 870.

The previous week's model had inherent Frozen-Graph Operations issue because of which is did not perform well when used with a larger dataset.So, I had to re-train the model. The newer performed well.

For Bounding Box generation, I have HTML5 canvas where in I draw the rectangle with the help of context2d. Total time for bounding box drawing and model inference is about 25 seconds for 3 min clip.

The video clip tested contains the images which are part of the segmented images obtained from the kavsir dataset. For joining the images I have used a software called as clideo.

The POC works well on Android phones without any major performance issues.

That's all for today!!!

Hope you had a great week

Au Revoir

References

Clideo

Stage 2 Assessments Capture with Classification

• Created stage-2 assessments capture with clasifying baby’s health using program rules
• Health is classified in three categories (Normal, Problem, Danger)
• App takes all data entered by user in stage 2 & then evaluate program rules for classifying health
• After classifying health the color of assessments screen changes according to the category

Stage 2 Notifications showing feature

• Used awesome_notifications package for displaying notification
• Displaying scheduled notifications reminder if user has not completed assessments
• Displaying notification on stage assessments opening
• Created Offline Notification showing functionality
  1. Showing notification when baby is registered to complete phase -1 assessments
  2. If user forgets to do phase 1 assessment then reminding by notification again after 60 minutes
  3. Notification feature works offline using scheduled notification featur
  4. Package used for showing notification https://pub.dev/packages/awesome_notifications

Screenshots of notifications

Unit testing assessments bloc, client, repository, phase-1

Test Cases

• Flutter Assessment BloC test
  • Test Cases
  • yield AssessmentsInitial on successful fetching of data
  • yields AssessmentsError on unsuccessful fetching of data
  • yields AssessmentsAdded on successful adding of phase 1
  • yields AssessmentsError when ward name is empty while adding of phase 1 assessments
  • yields AssessmentsError when assessments are not completed while adding of phase 1 assessments
    
• AssessmentsRepository testing
  • Test Cases
  • Throws exception if wardname is empty
  • Throws exception if assessments are incomplete
  • Marks phase-1 as complete if everything is correctly filled
  • Adds stage-1 assessments if list is empty
    
• Assessment Client Test
  • Test Cases
  • returns response with 200 code on successful fetching of assessments
  • throws fetch data exception on no internet connection

Searching & sorting functionality on list of babies screen

Sort List of babies by

• Time (Birth Time)
• Status ( Classificiation - Normal, Problem, Danger ) (Danger & Problem have high priority)
• Location ( Ward Name)

Search List of babies by mother’s name

What progress have you made this week?

• Adding Search, Sort Functionality in List Of Babies screen
• Unit Testing Assessments Bloc, Client, Repository, Phase 1
• Created Phase-1 Notifications showing functionality
• Created Phase-2 Assessments Capture with notifications

What do I plan to do next week?

• Phase - 3, 4 assessments capturing functionality
• Showing summary in Home Screen
• Developing settings screen of app
• Unit Testing phase - 2 , 3 , 4

Stage 1 Assessments Capture Functionality

• Fetching & Capturing Stage - 1 Assessments , with this feature user is able to view the captured assessments & also capture the assessments and push the data to DHIS2.
• This works offline also, data is sent to server when user refreshes on having internet next time
• Created a BLoC to handle assessments
• Created Stage - 1 model class which parses and converts to JSON format for parsing and converting data
• added required ID’s in static variables in dhis2_config.dart file
• created a client to send data, created a repository to handle data from client

Fetching list of babies enrolled in program

• Created a BLoC for fetching list of babies enrolled in a program
• Creating a client & repository class for fetching & saving data locally
• Sorting the list by time to categorize recently & past registered babies
• On Tap of baby list item the app opens assessments screen & start fetching assessments
• List item color changes with classification of baby - Danger (Red), Problem (Yellow), Normal (Green)

Merge Request Link

User Profile Screen Functionality

• Fetching list of users notification from dhis2
• saving data in hive storage after fetching item
• fetching & displaying the notification date & time for each notification

Merge Request Link

Localized strings in assessments phases UI

• All strings used in assessments phases UI are now localized
• localized strings are saved in app_en.arb file

Merge Request Link

What progress have you made this week?

• Implemented User Profile Screen functionality
• Implemented fetching list of babies from dhis2 & displaying in screen
• Cache the data after fetching from dhis2
• localized strings in assessments phase
• Phase 1 assessments capturing functionality

What do I plan to do next week?

• Phase - 2, 3, 4 assessments capturing functionality
• Adding Search, Sort Functionality in List Of Babies screen
• Adding sort functionality , settings UI, activity logs in profile screen
• Update functionality in profile screen

WEEK4

PROGRESS MADE THIS WEEK:

Advancing into the 4th week, everything is going as per the plan. By the end of this week, most of the components planned are already created.

This week (4th week) was utilised creating components for the visit workflow.

The visit workflow primarily involves the nurse and the provider along with the patient, the nurse checks the arrival of the patient, takes the vitals, verifies and creates medications and allergies.

Provider then checks the patient and performs exams and orders lab tests if necessary.

FHIR does not have a visit resource so every operation related to a visit event is handled by the modified encounter resource.

In FHIR, Appointment is used for establishing a date for the encounter,

When the patient arrives and the visit is about to start, then the appointment will be marked as fulfilled, and linked to the newly created encounter.

The following components were created for this: (other components which will be required apart from the following mentioned components are already created)

• Encounter
• Service Request
• Medication Statement
• Quantity datatype component
• Allergy intolerance
• Observation

The web component repo already had components for allergy and observation, but their utility was limited, moreover for every property there was a components which was actually not required as using created datatype components all the properties of any resource could be included in a single component.

A definte pattern is used for creating these components. The pattern is demonstrated below:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

/*imports required*/
import { LitElement, html, css } from 'lit-element';

class MyCustomElement extends LitElement {

  //all the properties are defined here
  static get properties() {
    return {
      value: { type: Object, reflect: true },
    }
  }

  constructor() {
    super();
    this.value = {}; 
  }
   
   // handle the change in input
   setDisplayValue(e) {
    this.value.display = e.target.value;
  }  

  // templates to render the component
  displayTemplate() {
    this.value.display = this.value.display || "";
    return html`<div>HELLO!</div>`
  }

  // render al the templates created 
  render() {
    return html`
      ${this.displayTemplate()}
    `
  }
}

customElements.define('my-custom-element', MyCustomElement);

Here are the screenshots:

—

—

next week:

• create the remaining component.
• implement visit workflows in the react application.

see you next week 😄.

WEEK4

This week was an eventful one. I started the week with TF Object Detection model development. I trained two models one was Faster RCNN and another one was MobileNet SSD. I plan to train YOLO models too and convert them into ONNX format for JS inference.

For VR the bounding boxes posed a lot of problems . For rendering the bounding boxes I used HTML5 Canvas and WebGL for designing and rendering the bounding box. The Dataset for Segmentation has only one class so class labels are not rendered on top of the bounding box. The model inference is still slow and in some cases fails to render the bounding boxes.

POC for Object detection is not working properly at this juncture. It still requires few more corrections for improving the model performance.

That's all for today!!!

Hope you had a great week

Vale

Hello everyone, do you want to know what is the progress of our project?😏

Obviously, I am here to share that with you. So let’s have a look at the work done in the fourth week of this amazing journey🧐-

This week I had set up Gitlab CI for our trainer app, Completed the offline support for the trainer app using SQLite database, added syncing UI in the app to differentiate the downloaded content, testing 2–3 issues that were mentioned in the last blog that they are resolved or not. Wait guys let’s discuss all this in detail😀-

Without CI/CD it difficult to check whether the particular code will build properly or not and we have to think about a local clone of the project or any other option to get its latest apk. To solve this I had set up the CI/CD Pipeline as well with the MR sent last week and in the one sent this week also.

Resolved Issues: #112

committed here - ce83405a 

Last run Pipeline: #327895681

CI is also added to offline-support MR.

Initially, in the last MR documents on the media page and every media file get downloaded and then played there is not any support for storing any type of data locally to the device to make it available offline.

Therefore I had added an SQL database to the trainer app to store data. This week I had completed my pending work related to this which was just started last week.

Now the updated list of documents get saved to the local database whenever a user visits the media page with an active internet connection and if a user visits this page in offline mode even in that case he is able to use the application in the same manner as it is in online mode, in this case, it is fetching data from a local database.

When a user clicks on any media file if it does not download it takes some time to get downloaded the first time and show a progress UI. Let’s suppose if it is already downloaded in that case it gets played instantly without waiting for a newer version but every time when user clicks on the media either it is already downloaded or not, it hits the API to update the media file. In offline mode, nothing update happens and it played the last cached version stored in the device and in online mode, it instantly plays the last cached version and hit the API to listen for any updates on this file.

SQL Database Schema & Logic

1. When the application launches it creates a media table with 4 columns-

• id- stores the id of the document with which it is stored on dhis2
• name- stores display Name of the document
• fileuri- will be not null only when the document is downloaded and saved on the device, it stores the local location of the document on the device and in other cases its value is null
• mimeType- it stores the content Type or media type of the document like application/pdf, video/mp4, etc. to play the media using the fileopener2 plugin

2. Logic for storing a list of documents of media page-

• when a user visits this page new list will be fetched from dhis2 and now we need to take care of 3 cases : 

(1) DELETE : Document ids not present in the new list but present in our local DB. These documents are deleted from dhis2 and need to be deleted from local DB as well.

(2) UPDATE : After completing the first step, local DB is left only with those ids which need an update. We just update the particular row of their ids of local DB with whatever new corresponding data we received.

(3) CREATE : Now those document ids of a new list that are not passed through step 2 are one of those of newly added documents. We just need to create a new document with their data in our local DB.

These steps will take care of all cases for making a list of documents available offline.

committed here: e8b93759

3. Logic for offline support of media files-

• Whenever the user clicks on the media file it gets downloaded and saved to the device local storage and its corresponding file location and mimeType gets updated to the fileuri and mimeType key of local DB for that document. And on the next time, a user doesn’t need to wait for it to get downloaded it gets played from the last saved location.
• I had used PERSISTENT storage and cordova-file-plugin to save the file to the device.

committed here: 65ddad02

Resolved Issues: #69

Corresponding MR: !195

Pipeline: #329574660

I had added 3 types of sync symbols ( Pending, Completed, Error ) in the media page list.

I had also added a progress bar that shows the ongoing sync state.

committed here: bf49a629

There are some issues whose testing I have to do last week. Let’s discuss them one by one-

1. Track app usage data ( #36 )

• We want to track the usage of our app and gets its report on dhis2 or tracker app.
• For that, the page visit is set up for all pages means whenever the user visits any page then it gets recorded on local device storage as a counter of visits to that page.
• Now the need is to shift this data to dhis2 via Web API in an App Usage Tracker Program. This part is not done.

2. Role-based restricted access for viewing tracked entities ( #44 )

• We want that the TEI should be visible only to the person who had created it other person visibility to this should be restricted.
• This part is not working and since we had updated the mHBS app from dhis2 after 3 years it has a lot of changes and the new codebase is completely in Kotlin and well architectured and segregated.
• Therefore this part need rework. This is not working in the new app. You can have a look at the latest mHBS tracker app codebase from here.

3. Need to create a new program stage in dhis2 ( #10 )

• This issue is also cannot be closed for now.
• There is no commit in the latest code that can solve this for now.
• Needs rework.

Works Done in the Fourth Week

1. Set up CI/CD to automatically build the apk and check for errors.
2. Added full offline support for the media page for both media files and a list of documents.
3. Added file synced based UI for each media file.
4. Tested 3 Issues.

Important Links

1. Merge Request: Having code related to full offline support & SQL implementation of media page.
2. Solved Issue: #69 #112
3. Gitlab Pipeline : #329574660 #327895681
4. Tested Issues : #36 #44 #10

Work to be done in the next week

1. Working on Issue #36 

• Shifting all app usage data from local storage to dhis2

      2. Working on Issue #44

• Restricting the access of other users for viewing tracked entities

      3. Working on Issue #10

• create a new program stage in dhis2

CHECK-IN WORKFLOW :

Valid search parameters for this search are: [_id, _language, _lastUpdated, active, address, address-city, address-country, address-postalcode, address-state, address-use, birthdate, death-date, deceased, email, family, gender, general-practitioner, given, identifier, language, link, name, organization, phone, phonetic, telecom]`

SEARCH PARAMS :

GET [base]/Patient?identifier=[value]

GET [base]/Patient?name=[value]

CREATE PATIENT :

POST [base]/Patient

GET PATIENT :

GET [base]/Patient/[_id]

EDIT PATIENT :

PUT [base]/Patient/[_id]

check-in.mp4 from Shashwat on Vimeo.

APPOINTMENT WORKFLOW :

SCHEDULE TAB

Valid search parameters for this search are: [_id, _language, _lastUpdated, active, actor, date, identifier, service-category, service-type, specialty]`

SEARCH PARAMS :

GET [base]/Schedule?actor=Practitioner/[value]
GET [base]/Schedule?actor=Location/[value]
GET [base]/Schedule?actor=Patient/[value]

GET [base]/Schedule?service-category=[value]

CREATE SCHEDULE :

POST [base]/Schedule

GET SCHEDULE :

GET [base]/Schedule/[_id]

EDIT SCHEDULE :

PUT [base]/Schedule/[_id]

SLOT TAB

Valid search parameters for this search are: [_id, _language, _lastUpdated, appointment-type, identifier, schedule, service-category, service-type, specialty, start, status]`

SEARCH PARAMS :

GET [base]/Slot?status=[value]
GET [base]/Slot?schedule=[value]

CREATE SLOT :

POST [base]/Slot

GET SLOT :

GET [base]/Slot/[_id]

EDIT SLOT :

PUT [base]/Slot/[_id]

APPOINTMENT TAB

Valid search parameters for this search are: [_id, _language, _lastUpdated, actor, appointment-type, based-on, date, identifier, location, part-status, patient, practitioner, reason-code, reason-reference, service-category, service-type, slot, specialty, status, supporting-info]”`

SEARCH PARAMS :

GET [base]/Appointment?actor=[value]
GET [base]/Appointment?status=[value]

CREATE APPOINTMENT :

POST [base]/Appointment

GET APPOINTMENT :

GET [base]/Appointment/[_id]

EDIT APPOINTMENT :

PUT [base]/Appointment/[_id] `

VISIT WORKFLOW:

///////////////////////////////////////////////NURSE///////////////////////////////////////////////

FIND PATIENT:

SEARCH PARAMS :

GET [base]/Patient?identifier=[value]
GET [base]/Patient?name=[value]

once the nurse selects a specific patient the patient Id will be
stored by a redux action. So the patient id below will be fetched from
the redux store via a useselector hook to get the appointments and the
encounters related to a specifc Patient entry.

APPOINTMENTS RELATED TO THE PATIENT:

GET [base]/Appointment?actor=Patient/[_id]

ENCOUNTER RELATED TO THE PATIENT:

GET [base]/Encounter?subject=Patient/[_id] /* TO GET ALL THE ENCOUNTERS SPECIFIC TO A PATIENT */

POST [base]/Encounter /* TO CREATE AN ENCOUNTER */

PUT [base]/Encounter/[_id]

ALLERGY RELATED TO THE PATIENT:

GET [base]/AllergyIntolerance?patient=Patient/[_id]

POST [base]/AllergyIntolerance /* TO CREATE AN ALLERGY ENTRY FOR THE PATIENT */

MEDICATIONS THE PATIENT IS TAKING/HAS TAKEN:

the fhir resource used here is MedicationStatement : this resource
has a reference to the patient, this resource is required to know what medications the
patient has been taking / taken before.

GET [base]/MedicationStatement?subject=Patient/[_id]

POST [base]/MedicationStatement /* TO CREATE A MEDICATION-STATEMENT ENTRY FOR THE PATIENT */

PATIENT VITALS:

IN FHIR THERE IS NO SEPARATE RESOURCE FOR VITALS , VITALS CAN BE TAKEN
AS AN ENTRY OF OBSERVATION RESOURCE UNDER THE CATEGORY OF VITAL-SIGNS

GET [base]/Observation?category=vital-signs&subject=Patient/[_id]

POST [base]/Observation /* TO CREATE AN OBSERVATION ENTRY FOR THE PATIENT VITALS */

PUT [base]/Observation/[_id] /* TO EDIT OBSERVATION ENTRY FOR THE PATIENT VITALS */

/////////////////////////////////////////PROVIDER/////////////////////////////////////////

FIND PATIENT:

SEARCH PARAMS :

GET [base]/Patient?identifier=[value]
GET [base]/Patient?name=[value]

once the provider selects a specific patient the patient Id will be
stored by a redux action. So the patient id below will be fetched from
the redux store via a useselector hook to get the
encounters related to a specifc Patient entry.

ENCOUNTER RELATED TO THE PATIENT:

GET [base]/Encounter?subject=Patient/[_id] /* TO GET ALL THE ENCOUNTERS SPECIFIC TO A PATIENT */

PUT [base]/Encounter/[_id]

similar to the above case once an encounter is selected its id will be stored via redux store

OBSERVATION RELATED TO THE PATIENT IN THE CURRENT ENCOUNTER:

GET [base]/Observation?encounter=Encounter/[_id]

POST [base]/Observation

ORDERS RELATED TO THE PATIENT IN THE CURRENT ENCOUNTER:*

GET [base]/ServiceRequest?encounter=Encounter/[_id]

POST [base]/ServiceRequest

E-PRESCRIPTION WORKFLOW:

////////////////////////////////////////////PRESCRIBER//////////////////////////////////////////

FIND PATIENT:

SEARCH PARAMS :

GET [base]/Patient?identifier=[value]
GET [base]/Patient?name=[value]

once the nurse selects a specific patient the patient Id will be
stored by a redux action. So the patient id below will be fetched from
the redux store via a useselector hook to get the
encounters related to a specifc Patient entry.

ENCOUNTER RELATED TO THE PATIENT:

GET [base]/Encounter?subject=Patient/[_id] /* TO GET ALL THE ENCOUNTERS SPECIFIC TO A PATIENT */

PUT [base]/Encounter/[_id]

similar to the above case once an encounter is selected its id will be stored via redux store

ALLERGY RELATED TO THE PATIENT:

GET [base]/AllergyIntolerance?patient=Patient/[_id]

MEDICATIONS THE PATIENT IS TAKING/HAS TAKEN:

the fhir resource used here is MedicationStatement : this resource
has a reference to the patient, this resource is required to know what medications the
patient has been taking / taken before.

GET [base]/MedicationStatement?subject=Patient/[_id]

MEDICATIONS PRESCRIPTION:

the fhir resource used here is MedicationRequest

POST [base]/MedicationRequest

CHECK-IN WORKFLOW :

Valid search parameters for this search are: [_id, _language, _lastUpdated, active,

This was the third week of the coding phase and it was good. This week majorly, I had worked on the way to play the video blobs that I had fetched from the server last week and showing the pdf files in the app. I had worked on the offline support of the application as well. Let’s see everything in detail-

Added a feature to play all type media files in app

We have 2 types of media files videos and pdfs which we have to show in our app’s media section. Let’s see what approach I had chosen to achieve that and why?

1. Firstly I had downloaded the blob object of the media that is being uploaded on the dhis2 using XMLHttpRequest.
2. Then I had created a local temporary space in the memory to save the file.
3. And lastly using the fileOpener2 plugin, I had opened that file using the supported apps that are already available on the user’s device.

This seems simpler but really It takes a lot of time to figure out how actually Cordova handles all this stuff🤔because I am not much familiar with the Cordova platform but finally I had made it. Actually, for videos I had initially used the most popular thing that is iframe but it didn’t work and I had changed it with the video element of HTML, and it gets working 😍. But the problem arises when I am trying to open pdf because in the browsers there are inbuilt plugins that handle pdf opening and all but for the android, there is no such thing therefore I need to use external apps with the help of fileOpener2 plugin. 

Final Output: Now one can upload pdfs and videos on the dhis2 resources and it will be fetched to our trainer app and can be accessed with the internet.

Have a look at the PR related to this task: Added mediaPage and Populated with real data Fetched from bmgfdev instance and Added support to play all media files of type blob (!194) 

Related Issues : #190 #77 #39

Added SQLite Database for storing data locally

We want our application to be active in offline mode as well therefore I had decided to use the sqlite database of the cordova plugin. Let’s look at the implementation part of the offline-support of media section:

 1. I had made a mhbsTarinerDb and media table to store the data of all the cuments. I will be storing a name (displayName of the document), id (dhis2 id of the document on the server), fileUri (location of the file if this media is downloaded). 

2. When the users click on the media tab then a list will be made from the documents available on the local db and it is shown on the tab at the same time an API call is initiated to the server if a user is online and we received a success response we will update our local db and media tab with this new data.

3. When a user clicks on particular media files then as stated earlier a blob object is downloaded and from it, we make a local file and save it to the device, now one point gets added that is we will store its location to our db as well and when the user again clicks on the particular files then similar thing happen means old content is shown till the new one is not downloaded and when new content is ready it gets updated to our db.

This part is almost ready and is under testing you can look at its progress from the offline-support branch of my forked repo. Worked done till now in offline support of media tab is here.

Work Done this Week

1. Added a feature to play all type of media files in the app.

• The list of all documents uploaded on the bmgfdev instance ( here ) is visible in the app.
• Videos Media & Pdf files both can be opened and can be accessed through the app.

      2. Added SQLite database to store data locally and to provide offline support. 

Planning of next Week

1. Finish pending work of Offline support and create PR for that.
2. Adding CI to Trainer App: Continuous integration through Gitlab CI (#112) 
3. Testing of some issues — 

• Role based restricted access for viewing tracked entities (#44) 
• Track app usage data (#36) 
• Need to create a new program stage in dhis2 (#10)

Important Links

1. Added mediaPage and Populated with real data Fetched from bmgfdev instance and Added support to play all media files of type blob (!194)
2. Issue #190
3. Issue #77 
4. Issue #39
5. Added SQL Database for storing data locally (26ad4006) · Commits · Bhavesh Sharma / mHBS-trainer · GitLab

Basic POC (Proof of Concept) design was completed and successfully implemented using WebXR and react. A JSON file was created which stores the links and names of all image file present in the directory.

The BiT model was converted using TensorflowJS model converter. The model then generated was used in POC. Images are cast on a Curved Plane in the VR environment, on which model inference runs.

Every 5 Seconds a new image is loaded and inference is given with 4 seconds. The labels are displayed as Text on top of the Plane in the VR environment.

The Video demonstrates a sample scenario. The inference reports undefined because the model was not able to classify the successfully

That's all for today!!!

Hope you had a great week

Do svidaniya

The previous model trained was not performing well on the the testing dataset. So, I had to try a new approach for increasing the model accuracy as well as other parameters. When I tried training with other pre trained networks, after one epoch model started to overfit. So, I had to get creative.

I Used a  BiT model for training, only the last layer was trainable. The model was able to perform better than pervious pre-trained model based on MobileNet.

Regarding the VR Part. The WebXR setup was successful. I was able to set up a basic WebXR environment. TensorflowJs was used for model inference. I used react-framework for binding TensorflowJS and WebXR together. For access to onboard device sensors the WebXR content has to be served through HTTPS.

Model Training Results

MobileNet Based Model: loss: 0.9046 - acc: 0.7174 - f1_m: 0.2450 - precision_m: 301254.4277 - recall_m: 0.2275 - val_loss: 1.0795 - val_acc: 0.6811 - val_f1_m: 0.2366 - val_precision_m: 248592.8750 - val_recall_m: 0.2270

BiT Based Model:loss: 0.6331 - acc: 0.8733 - f1_m: 0.8899 - precision_m: 195757.4552 - recall_m: 0.9773 - val_loss: 1.2562 - val_acc: 0.7955 - val_f1_m: 0.9079 - val_precision_m: 126641.6953 - val_recall_m: 1.0619

Inference Time:

When the SavedModel is converted into TensorflowJS model using the converter few seconds get added up in the inference time. So far, the inference time is not that much and is close if not less than 1 sec.

That's all for day!!

Hope you Had great Week

Ciao

Internationalized Flutter app by adding support for other languages

Created .arb (Application Resource Bundle) file which contains key-value pair strings, I localized all the string text using till now in app. For now I have added translations for 3 languages only

Added metadata of ECEB & on call doctors programs

DHIS2 provides a metadata repository containing metadata packages with various content. A metadata package is a DHIS2-compliant JSON document which describes a set of metadata objects. I exported this metadata for the ECEB & On call doctors program. Users who will use this app will need to import this metadata in their dhis2.

Unit testing for authentication screen

Unit tests are handy for verifying the behavior of a single function, method, or class. I created unit test for the auth functionality.

Test Cases for Hive Storage

1. User loggedIn return true
2. Not LoggedIn return false

Flutter BloC test cases

1. yield AuthenticationLoaded on successful login
2. yield AuthenticationError on unsuccessful login

Auth Client Test Cases

1. response code 200 - successful login
2. response code 401 - unauthorized (wrong credentials)
3. Socket Exception - no internet connection

Register Baby screen functionality

• Baby able to get registered using app & their data is stored using api in DHIS2.
• This feature works offline also, data is synced when user is online next time
• Created a refresh bloc for state management
• Added Snackbars for showing info related to syncing
• stored the list in hive storage

What progress I made this week?

• Internationalized Flutter app by adding support for other languages
• Added metadata of ECEB & on call doctors programs
• Unit testing for authentication screen
• Register Baby screen functionality (using DHIS2 API)

What do I plan to do next week?

• Implementing User Profile Screen functionality
• Baby Health Assessments Capture Functionality
• Fetching List Of Babies screen in App using API

Creating Login User functionality

• Used flutter Bloc architecture for state management
• Used DHIS2 Auth API for authentication
• username & password is passed as Basic Auth in header in base64 encoding in API
• Users can now login using username & password in App
• Added Dark mode support
• Created config files to store dhis2 apis & id
• Added password hide/show feature

Screen Shot

Created UI for Baby Assessments

Assessments are divided into multiple phases-

1. Phase 1 Assessments 0 - 90 minutes from birth

2. Phase 2 Assessments 60 - 90 minutes after phase 1

3. Phase 3 Assessments

Categorized in 3 parts

1. Normal
2. Problem
3. Danger

According to the Classification the appropriate assessment will be shown to user

Screen Shots

Added Theme for Dark Mode

Toggle feature is pending but I have enabled support for dark mode

Screen Shots

What progress I made this week?

• Completed authentication/login functionality
• Created config files for storing API, DHIS2 Id’s
• Developed UI of all three Assessments screens
• Added support for dark mode

What I have planned to do next week?

• Register Baby screen functionality (using DHIS2 API)
• Implementing User Profile Screen functionality (using DHIS2 API)
• Making the above features work also when user is offline and sending data when user is online next time

Hello Guys! The most exciting thing has begun that is the coding phase of GSoC-21. My community bonding period was gone good in discussing things and setting up the environment. This blog contains the work done by me in the First Week of the coding phase.

First of all, I want to share one important point with you :

1. We had shifted our mHBS-Tracker app to GitLab from Github. Its new link is here.

Merged latest changes of dhis2/dhis2-android-capture-app

Since our repo is very much lagging the new changes of the dhis2 app. The most important task that should be done at the first priority is to merge all the upstream changes. Our repo has approx. 1118 commits whereas in today’s time dhis2 app has approx. 8K commits. Therefore as a first task, I had merged all the upstream changes to our repo in order to work on the dhis2 app.

Modified the dhis2 app to work as mHBS app

After merging the upstream changes we are ready to work on the dhis2 app. Now to use it as a mHBS Tracker app we need to do some modifications to it as per our need. Therefore I had made all those UI changes and App Data changes to make it work as per the needs of the mHBS Tracker App.

List of modifications and other features :

1. App Name and Logo has been changed.
2. UI Improvements.
3. Added External others app of mHBS to app drawer.
4. Added functionality to open other mHBS apps.

Screenshots of the New mHBS App :

Setup CI/CD Pipeline to automatically build apk/artifact

Without CI/CD it difficult to check whether the particular code will build properly or not and we have to think about a local clone of the project or any other option to get its latest apk. To solve this I had set up the CI/CD Pipeline as well with the MR sent.

Fixed some of the known build errors

There are various build errors that were causing the problem in the successful build of the application I had solved a few of them. But still, there are some issues which need to be resolved but they are not interrupting the apk to build, these are just be reflected when apk is building. You can get a successful build and the latest apk from the code now as well.

Works Done in the First Week 

1. Merged the upstream changes that have more than 6K unmerged commits.
2. Modified the final merged code to have features and look like the mHBS app.
3. Set up CI/CD to automatically build the apk and check for errors.
4. Fixed Some of the build errors.

Important Links

1. Merge Request: Having code of all the work done in the first week. It has been merged to the main branch successfully.
2. Last Triggered CI Pipeline: This pipeline was triggered lastly in the MR.
3. Last Triggered Job and Artifact Link: Contains all the build log printed while running the ‘assembleDebug’ job. You can download the apk from the right panel of this job page from the ‘Job Artifact’ section.
4. CI/CD Setup Issue: Link of the CI/CD Setup issue solved with this MR.

Work to be done in the next week 

1. Solved the build errors of the mHBS Tracker app. There are mainly two problems for the mHBS app :

• Fix error ‘ > java.lang.NoClassDefFoundError: javax/annotation/Generated ‘ in gradle build when building with jdk 9+
• Fix Error in gradle build — string too large to encode using UTF-8 written instead as ‘STRING_TOO_LARGE’

2. Start working on the mHBS Trainer app for the media section.

3. Check for the complete smooth working of mHBS app and Improvements.

Developing UI of Register Baby Screen

I worked on creating UI of register baby screen, I have used the following packages

• Flutter Customizable Slider for weight, temperature input (Link)

• Flutter Date Time Picker to pick baby’s birth date & time (Link)

• Flutter Bloc Architecture to handle state management (Link)

I have created toggle buttons for baby’s information input such as gender, mode of delivery etc. These toggle buttons uses boolean variable for their state management.

I have created a model class for storing Baby’s health information, This class’s object is created & data is filled when user enter’s data is register baby’s screen.

Creating Navigation Drawer in App

I Added a navigation drawer in app, It has the following items & header with LibreHealth’s Logo.

• Doctor’s Schedule
• Messaging
• About
• Share App
• Report Bug

Screen Shots

What progress I made this week?

• Developed UI of Register Baby Screen
• Created Sliders for weight, temperature etc.
• Created Model Class for storing Baby’s Information
• Created Bloc for register baby screen
• Created App Drawer with items ( Doctor’s Schedule, Messaging, Share App, Report Bug)

What I have planned to do next week?

• Developing Assessments Screen UI for all stages
• Using Sliver App Bar for floating app bars
• Adding Dark Mode in App

This week was all about model building and model validation. Initially, I was stuck a bit on the metric portion, but was later able to overcome it by using special keras libraries. The labyrinthine keras and tensorflow documentation proved to be invaluable.

Model Development

Dataset

The dataset has labelled-images, unlabelled-images, labelled videos and segmented images. For the classification process I am going to use the labelled-images. They have 22 classes labelled in them.

Model

I am training two model, out of which one model will be a custom model and other one will be pre-trained one which is adapted to Kvasir Dataset. A size of 224x224 will be used as the input shape for the model. For data logging purpose i have used tensorboard.

Results

Both the models were trained. The pre-trained model provides low generalisation error in-comparison to the Custom-Model.

CustomModel: loss: 2.4398 - accuracy: 0.7283 - f1_m: 1.1505 - precision_m: 142484.9571 - recall_m: 1.1362 - val_loss: 3.0672 - val_accuracy: 0.0990 - val_f1_m: 0.9958 - val_precision_m: 0.9958 - val_recall_m: 0.9958

PreTrained Model: loss: 0.9046 - acc: 0.7174 - f1_m: 0.2450 - precision_m: 301254.4277 - recall_m: 0.2275 - val_loss: 1.0795 - val_acc: 0.6811 - val_f1_m: 0.2366 - val_precision_m: 248592.8750 - val_recall_m: 0.2270

Virtual Reality

For handling virtual reality portion of the project I am going to use WebXR. This doe not require installation of application on the phone. It supports various platforms like oculus and android

Model Architecture

That's all for today. Meet y'all later

So long

Tl;dr: Model dev cycle complete

References

MobileNetV2

WebXR

CustomModel

PreTrained

Community Bonding Period June 1 to June 6

Coding period is about to begin tomorrow & I have done preparations to start coding, During this week I had completely set up DHIS2 with program rules/Stages

Few thing I want to Discuss

Baby Classification Logic

Earlier I thought to do Classification in app only , but now I learned about program rules which run automatically and can classify baby easily, Moreover these rules are very much configurable so users can configure these according to their requirements

Currently I used ECEB Action plan to classify Baby’s health in 3 categories - Normal/Problem/Danger

Example of Classification Logic For Danger

#{ECEB_Severe_Jaundice} || #{ECEB_Assess_Temperature}<= 35.5 || #{ECEB_Assess_Temperature} >= 37.5 || #{ECEB_Weight} < 1500 || #{ECEB_Chest_Indrawing} ||#{ECEB_Not_Feeding} || #{ECEB_Fast_Breathing} || #{ECEB_Convulsions}

If any of the above condition is true then the health is in danger.

The program rule has an acion - It classifies health by filling data element ECEB_Classification after evaluating the above expressions.

Creating an Initial Screen to fetch all data elements/program ids from dhis2

For doing API call related to a program/event we need its unique id beforehand, As these ids are not constant on every DHIS2 , I need to first design a screen which fetches all the variables/Stage/Programs ID from the dhis2. Then the user will be able to use the app.

On Call Doctor’s Program

This is an Event Program, Doctor’s schedule is added as an event with data elements name, date, start time, end time, Doctor’s Image.

What progress I made this week?

• Created program rules for Normal/Problem/Danger
• Created On Call Doctor’s Program
• Created Stage 4 & Stage 5 in DHIS2 & modified previous stages according to ECEB action plan

What I have planned to do next week?

• Developing Register baby screen UI
• Developing Assessments Screen UI of App
• Adding Dark Mode in App
• Adding navigation drawer in app with menus ( Doctor’s Schedule, Messaging, Share App, Report Bug)

 Hello everyone, I am glad to share that my project proposal was accepted by LibreHealth in GSoC-21 and I am going to spend my summer working with LibreHealth Community on the Project: Scaling up the mobileHBS/DHIS2 Tracker and Trainer applications — Projects / mHBS — LibreHealth Forums.

I am writing this blog to share my community bonding phase with LibreHealth. It was scheduled from 17th May to 7th June and the coding period will begin in about 10 days.

Project Description

The mobile Helping Babies Survive powered by DHIS2 (mHBS/DHIS2) Tracker (Java) and Trainer (HTML; JavaScript) applications, built within the DHIS2 Mobile platform, are designed to provide educational and training support for maternal and neonatal care among health care providers in low/middle-income countries. We help nurses and physicians in sub-Saharan Africa, Asia, and Latin America deliver evidence-based interventions that save maternal and newborn lives.

In various low/middle-income countries children die at a very low age just in few days after their birth due to the inadequate supply of facilities that they require. The important fact is that all of these deaths are PREVENTABLE and can be prevented by providing proper knowledge to the mother’s of the babies as well as to other person and by providing a proper tracker facility to track the health status of the baby in the early days of his birth.

mHBS (mobile Helping Babies Survive) solve these issues. There are 4 mobile applications, in various stages of design, development, evaluation, and deployment, within the award-winning mHBS/DHIS2 suite of digital tools.

1. mHBS/DHIS2 Tracker (work: data collection)
2. mHBS/DHIS2 Trainer (work: educational and training materials)
3. Essential Care for Every Baby Digital Action Plan (work: clinical decision support)
4. NeoRoo (work: data collection, education, and automated monitoring of neonatal vital signs among babies utilizing the NeoSMART biomedical device)

I will be working on the scale-up of the already existing mHBS/DHIS2 Tracker and Trainer app so that they can be published on the Play Store.

Community Bonding Period

I had used the last days of my community bonding period to set up the project and in getting familiar with its code base and most of the time I did one thing that was getting more idea of the project i.e. Research… Research… Research… and getting familiar with technologies that are going to be used in the coding phase to write a structured code.

We are going to use the dhis2 app codebase for our mHBS Tracker app to collect data from the users and mHBS Trainer app is not a java or android project it is written in HTML, CSS & JavaScript and we are using Cordova for running it on Android Platform.

1. Read articles, documentation related to the project.
2. Learned how to use Cordova for running it’s project on Android.
3. Set up the codebase and ready to code.
4. Learned a lot about the technologies and codebase.

Important Links

1. Talk thread: Proceeding With GSoC
2. Trainer App: GitHub Repo
3. Tracker App: GitHub Repo

Conclusion

It was very great to start with the LibreHealth community.

This week onwards, I will be posting regular weekly blog posts on the updates on further progress on this project here. Thanks for reading! 🙂

Community Bonding Period May 18 to 1 June

Hello Everyone,

I am very happy to start my Google Summer of Code journey with LibreHealth, I will be working on Mobile App Development Project, I am very much Thankful to LibreHealth & Google for selecting me again. I have also successfully completed GSoC previous year & have worked on Cost of Care project of LibreHealth.

About Project

To develop an Android & IOS application to provide clinical decision-support for nurses and doctors delivering essential newborn care interventions during the first day of life. This application provides knowledge, skills, and competencies to nurses and doctors in low/middle-income settings so that they can provide life-saving care to newborns from birth through 24 hours postnatal. App uses DHIS2 as backend & can be setup for any hospital/organization

The Goals of this project are

• Ability to track multiple babies at a time;
• Time-stamping births to generate an essential newborn care clock for each baby being tracked in the app;
• Automated classification of babies health status, based on results from observations and assessments that are conducted during the first 2 hours after birth;
• Automated advice on management of babies, based on their classification; Reminders/alerts to provide recommended essential newborn care interventions;
• If particular interventions are not delivered on time, or at all, capture of information regarding why this essential newborn care was late, or missed

About Community Bonding Period

• Community Bonding Period in which you get to know your community and get familiar with their code base and work style.

• Ask for resources, required by you, to gain your knowledge for the parts of your project you don’t know.

• Discuss and plan your GSoC work with your mentor.

• Discuss and plan your time-table for your summer.

My Progress

• During the community bonding I have introduced myself to mentors & mentors are very much welcoming.
• I recieved access to DHIS2 for the application development, I have discussed with them about creating Programs, Stages, Attributes in DHIS2.
• I read the documentation of DHIS2 Tracker API, Creating Program/Stages.
• I have created a program for On-Call-Doctors & I have to discuss it with mentors.
• I resolved my doubts related to modifying the already created program in DHIS2.

Tasks For Next Week

• Completely setting up DHIS2 with all Programs/Stages/Attributes
• Finalising weekly tasks & creating milestones
• Reading Documentation
• Discussing & showing mentors the programs I will modify in DHIS2

Hi My Name is Shivaditya Shivganesh, This is first blog post with regards to Google Summer of Code commonly known as GSoC for the year 2021.

I have been selected for the  Project: Neural network-based object detection of anatomical structures and medical artifacts in Virtual Reality

At present we are in the community bonding phase. Which last for about 15 days the actual coding portion of GSoC starts from 7th of June.

I have started to work on Neural Network model that would be required for the project.

About the Project

In this project we are trying to create a Neural Network model that can. classify the respective classes in the Kvasir dataset and display it in an appropriate VR Display.

The Hyper-Kvasir dataset consists of  images which can be classified into 23 classes.

Conclusion

This blogpost was just for the  community bonding purposes. In the forthcoming weeks I plan to write posts regularly about my progress and the challenges I faced.

Until next time

Hi My Name is Shivaditya Shivganesh, This is first blog post with regards to Google Summer of Code commonly known as GSoC for the year 2021.

I have been selected for the  Project: Neural network-based object detection of anatomical structures and medical artifacts in Virtual Reality

At present we are in the community bonding phase. Which last for about 15 days the actual coding portion of GSoC starts from 7th of June.

I have started to work on Neural Network model that would be required for the project.

About the Project

In this project we are trying to create a Neural Network model that can. classify the respective classes in the Kvasir dataset and display it in an appropriate VR Display.

The Hyper-Kvasir dataset consists of  images which can be classified into 23 classes.

Conclusion

This blogpost was just for the  community bonding purposes. In the forthcoming weeks I plan to write posts regularly about my progress and the challenges I faced.

Until next time

Hi My Name is Shivaditya Shivganesh, This is first blog post with regards to Google Summer of Code commonly known as GSoC for the year 2021.

I have been selected for the  Project: Neural network-based object detection of anatomical structures and medical artifacts in Virtual Reality

At present we are in the community bonding phase. Which last for about 15 days the actual coding portion of GSoC starts from 7th of June.

I have started to work on Neural Network model that would be required for the project.

About the Project

In this project we are trying to create a Neural Network model that can. classify the respective classes in the Kvasir dataset and display it in an appropriate VR Display.

The Hyper-Kvasir dataset consists of  images which can be classified into 23 classes.

Conclusion

This blogpost was just for the  community bonding purposes. In the forthcoming weeks I plan to write posts regularly about my progress and the challenges I faced.

Until next time

Hello !

Its been quite a while since the GSoC results were announced and community bonding period commenced. The calendar says May 30 and the coding period will begin in about 10 days. So lets move on to how I actually spent the community bonding period yet and what plans do I have for the coming days before the coding period starts.

Project Introduction

The project bases its goal on creating an EHR system using the webcomponents implementing FHIR resources developed during the previous versions of GSoC. For the implementation some new resource based components will be required as well which will be created simultaneously.

During GSoC 2018, under this project a polymer application was developed which combined all the web component0s (developed at that time) into a PWA.

The current project will be built upon this application by adding and implementing workflows. the workflows that will be implemented in the EHR are :

• Check-in
• Visit
• e-Prescription
• Appointment
• Lab orders

Amazing Mentor Team

1. Saptarshi Purkayastha
2. Namratanehete

Community Bonding Period

Background

During the application period upon the suggestions of the mentors I had decided to mainly work on the CI and the Designs of the PWA during the community bonding period.

Even before the results were announced I was already experimenting with setting up the project so as to identify any problems that may occur as using lit-components inside React can be a little trouble because React never exposes the real DOM.

Bonding Period

So as soon as the results were released I started looking on these parts. I started with the CI part first, one of the pointers in this was also to look for a cross platform - cross browser testing method (SauceLabs) suggested by my mentor. I looked at some of the possibe integrations with these automation tools here.

But most of these are based on e2e testing ( cypress , Puppeteer and others ) and our main focus is on unit and integration testing which will require the react component to be mounted and tested , for which many tools are available but I could not find their integration with sauce or any cross browser-cross platform tool. Once I figure out a way for this issue the CI part will hardly take any time !

the design part:

I have designed one workflow ( check-in ) and will complete the remaining in the coming 2-3 days ! adding to these I will document the endpoints that will be used in each steps in the designed workflows. The endpoints of the check-in workflow is provided below.

here is the video of the designed check-in workflow:

check-in.mp4 from Shashwat on Vimeo.

CHECK-IN WORKFLOW :

SEARCH PARAMS :

GET [base]/Patient?identifier=[value]

GET [base]/Patient?name=[value]

CREATE PATIENT :

POST [base]/Patient

GET PATIENT :

GET [base]/Patient/[_id]

Conclusion

This week’s blog was not too long as I have not written any code, but hoping to deliver some interesting content in the upcoming weeks! I will be posting a blog every week highlighting the things I completed during the week and the things that I plan to achieve.

See you next week 😄

"Experience"

 This week I was successfully able to make my models produce good results. I was evaluating them on the wrong structure of data. Yes, yes the data format that I was working on since ages did not prove useful to me. But it was a building block to the generators that I finally created. After that, I re-evaluated my original model and viola! I got an AUC-ROC score of 0.81. As this generator model proved useful, I immediately parallelized my operations. 

I was running a total of 6 notebooks. 2 on PLHI server, 2 locally, 1 Kaggle notebook and 1 Colab notebook. 

The PLHI server did not have the complete data so I exported a subset over there and re-trained my pruning model.

My local system had the data but lacked processing power. So I ran model evaluations locally. Each Int-8 model took over 24 hours. 

The Kaggle notebook had data and processing power but the data was not in the correct structure to be fed to models. So I tested inference scripts here.

Colab lacked the data but had processing power. It also had support for Tensorflow-model-optimization library. So I ran my model quantization scripts here by uploading a small subset of the data.

The main problem was that the data is 45GB in size. It took a lot of time to download or upload anywhere. Even for my local system, I had to clear a lot of my personal data to make space for this dataset.

Along with the notebooks that I was running, I was also testing my custom scripts locally. Finally, I finished these processes and pushed my code. I also submitted my model results.

While I was working on these tasks, I also setup my Raspberry Pi on Qemu and began working on it. There were 2 main blockers here - network and space. The emulator was not getting internet access initially. So I built a bridge interface from my host ethernet to the guest (RPi) ethernet. This took quite some time to troubleshoot and in the end I was able to provide internet access to my Pi. A few tweeks later, I was able to get it to connect to my Wifi connection. 

The next blocker was the size of the image. After emulating, I was not left with enough space on my partition to install Tensorflow or other packages. So I had to resize the Raspbian image. This needed me to tinker with the partitions using fdisk and in the end, I was able to allocate space to the image.

Right now, I am trying to solve the MemoryError that I get everytime I  try to install Tensorflow on the Pi. I also get an error that says that grpcio is not supported. Hopefully this will get resolved soon because the scripts are ready to be run.

After this last task I can say that I have completed my project.

Happy Coding :)

The Beginning

From the beginning of my ‘Outreachy’ internship I had in mind that I was here to learn and to contribute the little I know. It was not easy regarding the fact that I had no working experience and most of all, it is an open source contribution. Following my core values and my background(https://ahanipeitra571448494.wordpress.com/2020/06/02/my-outreachy-participation-3/). I was able to pull through the various stages of the internship with the help of the understanding mentors I have in ‘LibreHealth'(https://librehealth.io/)

The Issues I faced

Writing exams and working with real life projects are purely different things. I came to realize this when i was selected as a LibreHealth intern.

I have gotten stuck with many issues over the period of my internship, some more tougher than others. For example, I got stuck on Mysql not running on my server.

Most of these issues should not have taken more than a couple of hours to figure out but more often it took me days. This is because I seek the help of my mentors mostly after I have tried figuring out my self. It was normal for me to think that the bugs that I was getting were because of a bigger task or things that I did not understand yet. This would force me to go extra miles to better understand what is going on. I will later find out at times that the issue I was facing was just minor or small mistakes I made. How ever I got more excited at the end because being stuck on something for long and finding the solution gives some one the mastery of it. The more difficulties I faced the more knowledge I accumulated and the more confidence and experience I gained.

I also faced the issue of getting the wright location of the files. Sometimes getting a particular line of code to work with is very tricky.

Some time ago it was ‘GitHub’ I could not get my branches or do other operation I spent days on this. It helped me to understand ‘GitHub’ better. From this link you can read about how I got stuck (https://ahanipeitra571448494.wordpress.com/2020/06/11/how-i-got-stuck-during-my-outreachy-internship/)

My Progress So Far

Though facing the issues I talked about I have been able to progress in both my work in the internship and my skills as a software developer. I am now able to work on an open source contribution and also on my own projects.

Me working with Librehealth As an Outreachy intern has also given me more skills and understanding on how to write a report and a blog. Before the internship I can recall how difficult it was for me writing reports in a good format and a blog but with the help and the guidance of my mentors I am now better on it.

I have been able to make some fixes in the ‘Libreehr System’ though am still on some few tricky issues to fix, but I believe I get more better with time.

Working as an Outreachy intern, I have also been able to improve on my English writing skills and communication. This because My mentors always have the patience and time to correct the mistakes I make and tell me what to do best. English writing and communication is one of the main things am so happy to have improved in.

Also getting to understand how to use Git’, ‘GitHub’, ‘WordPress’, editor like Visual ‘Studio code’ is a very big achievement for me as a software developer.

What I want After My Internship

From all the skills I have gathered during the Outreachy intership which is an open source contribution, I will love to continue contributing in an open source to better my skills. After that I will love to be an entrepreneur.I want to struggle and build my own web applications and also help my community.