This repository contains the Pediatric Apple Watch Study (PAWS) application. The Pediatric Apple Watch Study Application uses the Spezi ecosystem and builds on top of the Stanford Spezi Template Application.

The Spezi Template Application uses a modularized structure using the Spezi modules enabled by the Swift Package Manager.

The application uses the FHIR standard to provide a shared standard to encode data exchanged between different modules.

The PAWS app uses a Google Firebase-based backend. You will need to install the Firebase Local Emulator Suite to run a version of the service infrastructure locally. To boot up the Firebase backend, you need to run the following command to start the Firebase emulator, including a seeded version of the infrastructure. This is also required before running the automated user interface (UI) testing suite.

firebase emulators:start --import=./firebase

You can then visit http://127.0.0.1:4000/firestore/default/data/invitationCodes to see all the seeded invitation codes that can be used in the application. The PAWS app automatically connects to the local emulator when running in the iOS simulator using Xcode.

You can build and run the application using Xcode by opening up the PAWS.xcodeproj.

The ECG Data Manager provides capabilities for reviewing and exploring the recorded ECG data. It relies on the spezi_data_pipeline package, which is a library engineered to improve workflows associated with data accessibility and analysis in healthcare environments. In addition to the functions and classes of the spezi_data_pipeline, the two notebooks, namely ECG Reviewer.ipynb and ECG Explorer.ipynb, offer an environment for interactive data visualization and analysis.

The ECG Data Manager includes:

• utils.py: Provides utility functions for data processing.
• visualization.py: Contains functions for data visualization.
• ECG Reviewer.ipynb: An interactive notebook for loading, analyzing, and reviewing ECG data.
• ECG Explorer.ipynb: An interactive notebook for loading, exploring, and filtering ECG data based on filters, such as age group, ECG recording classification, user, and date.

You can open and run the ECG Reviewer.ipynb and ECG Explorer.ipynb notebooks in, e.g., Google Colab. Once the notebook is open, execute the following cell to clone the PediatricAppleWatchStudy repository and navigate into the cloned directory:

# Clone the repository and navigate to the ECG data manager
git clone https://github.com/StanfordBDHG/PediatricAppleWatchStudy.git
cd PediatricAppleWatchStudy/ecg_data_manager

To run the notebooks, add them to Colab Enterprise within the same Google Cloud project as your Firebase setup. For other Python notebook environments, use the Firebase credentials and upload the serviceAccountKey_file.json to the workspace directory to enable Firebase access. This file is essential for authentication and should be securely handled.

To start reviewing ECG data, execute the cells in your notebook.

This interactive tool allows you to plot ECG data, add diagnoses, evaluate the trace quality, and add notes.

To start exploring ECG data, execute the cells in your notebook.

This interactive tool allows you to plot ECG data, filter ECG recordings, and select specific users and timestamps.

The Google Cloud Setup at Stanford to deploy the project requires the following setup for Google Cloud Firebase and to execute the ECG review and exploration tools.

Each Firebase Project for development, staging, and production GitHub environments needs the following configurations:

• Firestore Database
• Firebase Authentication with Identity Platform (Anonymous Authentication, Username + Password, and Sign In With Apple enabled)
• Firebase Storage
• Cloud Functions (also enable Cloud Build API, Google Cloud Run, and Eventarc API)

The CI setup requires a github-deployment@PROJECT_ID.iam.gserviceaccount.com account that requires the following roles:

• Cloud Datastore Index Admin
• Cloud Functions Developer
• Firebase Admin
• Firebase Rules Admin
• Service Account User on [email protected]

Set up a cloudfunctionsserviceaccount@PROJECT_ID.iam.gserviceaccount.com to execute cloud functions. It needs the following rules:

• Cloud Datastore User

When signing in to the application for the first time, users will be required to enter a valid invitation code before a user account is created. Use the upload_codes.py script to generate new codes and upload them to a specified Firebase instance once you have deployed the firebase setup in this repository to your Firebase instance.

export FIRESTORE_EMULATOR_HOST="localhost:8080"
export GCLOUD_PROJECT=<project_id>

python -m scripts.upload_codes --outfile=<local_path> \
--count=<number_of_codes> --length=<code_length> \
--service_account=<service_account_key_file> [--dry]

Create a storage bucket that is used to store the packaged dependencies and code for the Python notebooks in a versioned and isolated state. You need to enable Colab Enterprise.

To secure the data in the notebook, the network access should be restricted to needed Google services only and use runners that only use a private network.

Run the following commands in the ecg_data_manager folder to package the dependencies and upload them, as well as the modules folder, to the cloud storage bucket if no outside internet access is enabled.

mkdir packages
pip download -r requirements.txt -d packages
tar -czvf packages.tar.gz packages/

Copy the ECG Explorer and ECG Reviewer notebooks in Colab Enterprise, uncomment, and adapt the storage bucket paths if restricted network access is configured.

The project supports different GitHub environments (development, staging, and production).

• The Firebase project ID needs to be saved as a GitHub variable with the name FIREBASE_PROJECT_ID for the different deployment environments.
• The service account key needs to be added to the GitHub secrets as GOOGLE_APPLICATION_CREDENTIALS_BASE64 in a base64 encoding to enable the beta deployment.
• To report code coverage, a CodeCov token should be added as a CODECOV_TOKEN environment secret.
• The Firebase Google plist needs to be stored as a base64 encoded secret named GOOGLE_SERVICE_INFO_PLIST_BASE64.
• Store all secrets for a beta deployment of the iOS application as documented for the Stanford Spezi Template Application.

PAWS uses Fastlane Snapshots to automatically screenshot specific screens in the app during UI tests. To generate new screenshots, you will also need to have a properly seeded local Firebase instance up and running:

firebase emulators:start --import=./firebase

Then, run fastlane snapshot. By default, results will end up in the .screenshots folder, overwriting previous files.

The current workaround for simultaneous account registrations during fastlane snapshot is to continually reset invitation codes to an unused state in Firestore by running a designated Python script on repeat (in a shell session with the same environment variables).

for i in {1..360}; do python -m scripts.upload_codes; sleep 10; done

Contributions to this project are welcome. Please make sure to read the contribution guidelines and the contributor covenant code of conduct first.

This project is licensed under the MIT License. See Licenses for more information.

For more information, check out our website at biodesigndigitalhealth.stanford.edu.