AURA Epileptic Seizure Detection

This repository contains everything required to run the AURA Seizure detection ML process on ECGs.

It notably features:

• Pipelines for data cleaning and preparation, model training, prediction and visualisation.
• An Airflow orchestration setup to run the various pipelines locally via a local executor.
• A Graphene orchestration setup to run the pipelines in a kubernetes cluster.

Introduction

AURA

Aura is a not-for-profit organisation working on epileptic seizure detection.

The AURA workflow basically relies on a RandomForest Machine Learning model to detect epileptic seizures from ECGs data.

Running with Airflow

TBD

Running with containers

See README.md in graphene/.

Contributing

We try to keep the repository clean and well maintained. The following rules apply:

• The feature works and has been tested on a CI system.
• There are tests to run every new or updated code. Tests must be documented, reproducible and automated as much as possible. We use pytest, but as long as the user can easily execute the tests easily it's fine.
• Python code follows the default configuration of flake8. One can use black to automatically reformat Python files.
• Documentation is present and current. If a new feature is introduced it must be documented, and for a change to an existing feature the current documentation must be updated.

We use Travis for all Python code, and container-based tests are executed on a Jenkins instance.

Pull Requests shall be accepted only if they meet all the above requirements.