ECG Classification with a 1D CNN

A 1D CNN that flags abnormal ECGs at 0.98 AUC and uses Grad-CAM to show which parts of the heartbeat drove each call.

• Trained a PyTorch 1D CNN for ECG abnormality classification, achieving an AUC of 0.98 in first-pass modeling.
• Used Grad-CAM to explain ECG predictions clinically, identifying decision-driving ST-segment and QRS complex regions.
• Optimized model performance with Optuna hyperparameter search over dropout, batch sizes, and learning rate scheduling.