DeepBeat: Multi-task deep learning for cardiac rhythm detection in wearable devices

DeepBeat is a deep learning model for
1) classification of Atrial Fibrillation (AF) from photoplethysmography (PPG) signals and
2) signal quality estimation

Dataset

We share a deidentified train, validate and test partition that were used for training DeepBeat. These deidentified signal are shared with a non-commerical data use agreement and can be found at synapase, ID:syn21985690.

Examples

We have included juypter notebooks and trained models for reproduciblility.