PROPHECG-Age

PRediction Of PHenotypes using ElectroCardioGraphy-Age
Goal : Predict ECG-age using raw ECG waveform and deep neural network

ECG Data processing

• Raw ECG waveform with 500Hz / 10 sec / 12 leads
• Use only 8 leads (I, II, V1-V6) as other 4 leads are computed using these 8 leads
• Input shape : (5000, 8)

Model Train & Test

Model architecture

Below image shows the architecture of 1 dimensional residual block neural network based convolutional neural network used for the age prediction. As the purpose is predicting age, this is a regression task. Final output after dense layer would be AGE.

Ref) Lima et al., Nat Commun 12, 5117 (2021). https://doi.org/10.1038/s41467-021-25351-7.

Folder contents

• train.py: Script for training the neural network. To train the neural network run:

$ python train.py SCRIPT_YAML

• evaluate.py: Script for generating the neural network predictions on a given dataset.

$ python evaluate.py SCRIPT_YAML VALID_DATASET

• resnet.py: Auxiliary module that defines the architecture of the deep neural network.

• CustomDataset.py: Customed Dataset to be put into DataLoader.

• script.yaml: Script with parameters needed for training and validation.