mandrakedrink / ecg-synthesis-and-classification

1D GAN for ECG Synthesis and 3 models: CNN, LSTM, and Attention mechanism for ECG Classification.

ECG-Synthesis-and-Classification

1D GAN for ECG Synthesis and 3 models: CNN with skip-connections, CNN with LSTM, and CNN with LSTM and Attention mechanism for ECG Classification.

Motivation

ECG is widely used by cardiologists and medical practitioners for monitoring the cardiac health. The main problem with manual analysis of ECG signals, similar to many other time-series data, lies in difficulty of detecting and categorizing different waveforms and morphologies in the signal. For a human, this task is both extensively time-consuming and prone to errors. Let's try to apply machine learning for this task.

Data

Available here.

Formulation of the problem:

Each signal should be labeled as one of the classes ("Normal", "Artial Premature", "Premature ventricular contraction","Fusion of ventricular and normal", "Fusion of paced and normal").

Solution

Code with research and solution is available here - 1D GAN for ECG Synthesis and here - ECG Classification | CNN LSTM Attention mechanism.

Models

GAN Results

Classification Results

ecg-synthesis-and-classification's Issues

Pre-trained model

Hi, would you share a pre-trained model to run the evaluations? Thanks!

Is there a paper with this code?

ECG real and synthetic

Hi. I'd like to ask how did you present each of the class real signal and its counterpart synthetic signal? Can I ask for a code? Thank you.

Accuracy problem

Hello, why is the accuracy of verification set greater than that of test set in this project code? And I use my own data set to verify that the accuracy of the set will be greater than 1, which is puzzling