search

DeepPSP / torch_ecg

Deep learning ECG models implemented using PyTorch
MIT License
161 stars 20 forks source link

Training with peaks #14

Open mario-sanz opened 6 months ago

mario-sanz commented 6 months ago

Hello,

I would like to train a UNET network with the LUDB dataset but also with the peaks, not just the P, QRS and T waves (I would like also to detect P, R and T peaks). Is this possible? I have been doing lots of tests but I am not able to do so.

Thanks,

Mario

wenh06 commented 6 months ago

Peaks are almost the "argmax" of corresponding waves, perhaps you can do a peak detection (sometimes valley detection) within the ranges of the waves after wave delineation (segmentation). Or you can narrow the segmentation labels to a small neighborhood of the peaks (e.g. ±20ms) and train a model using these labels, then the peaks would be the center of the segmentation result of the model inference.