🫀 AI Heart Murmur & Arrhythmia Detection Model
Trained on phonocardiogram recordings from the Kaggle Heartbeat Sound Challenge (abdallahaboelkhair/heartbeat-sound).
Model Architecture
- Bidirectional LSTM (BiLSTM) with Temporal Attention Pooling
- 60 Acoustic Features (40 MFCCs, 7 Spectral Contrast, 12 Chroma, 1 RMS Energy)
Target Classes
- Normal: Regular S1/S2 lub-dub heart rhythm
- Murmur: High-frequency turbulent systolic/diastolic noise
- Extrasystole: Premature cardiac contractions / arrhythmia
- Downloads last month
- 6