mit-bih

Machine Learning on ECG to predict heart-beat classification.

An explainable deep learning system for automated ECG arrhythmia detection using a hybrid 1D CNN–LSTM model with Grad-CAM–based clinical interpretability.

Newton–Puiseux for CVNNs: complete toolkit for uncertainty mining, confidence calibration and local symbolic-numeric analysis on ECG (MIT-BIH) and wireless IQ data (RadioML 2016.10A).

Production-ready ML system for automated ECG arrhythmia classification using MIT-BIH database

Arrhythmia detection using topological data analysis in combination with a convolutional neural network.

Chance-corrected benchmark of ECG representations (raw, autoencoder, hand-crafted, foundation models) for unsupervised arrhythmia clustering on PTB-XL and MIT-BIH, with a supervised ceiling and deployment-robustness (imbalance, federation) analysis.

Low-cost real-time ECG monitor using ESP32 + AD8232. FIR/IIR digital filters designed in MATLAB, implemented in C++ firmware. Detects Tachycardia, Bradycardia and irregular R-R intervals.

Deterministic ECG codec — Python + Rust, CI parity-gated. Bounded clinical-mode contract: PRD ≤ 2.32% on MIT-BIH (48/48, mean PRD 1.12%); PTB-XL boundary disclosed (max PRD 5.29%). Cardiologist-equivalence and regulatory closure out of scope.

An investigation into tabular classification with deep NNs for ETHZ Machine Learning for Healthcare on the MIT-BIH arrythmia dataset .

Explainable ECG arrhythmia detection using 1D-CNN + SMOTE + SHAP on the MIT-BIH dataset | Deep Learning | XAI | Medical AI

Containerized ML inference service for ECG arrhythmia classification on GCP Cloud Run, with a 1D CNN trained on MIT-BIH and GitHub Actions CI.

Five-class ECG heartbeat classification (MIT-BIH) with a feedforward MLP - imbalance-aware, 90.9% accuracy / 0.724 macro-F1. Notebook, trained model, figures and full report.

Heartbeat arrhythmia classification with a neural network written from scratch in NumPy — forward/backward propagation, optimisers and regularisation by hand, verified against PyTorch to 1e-17. Patient-disjoint evaluation on MIT-BIH, 216 logged experiments, and a documented model-selection failure.

MIT-BIH Arrhythmia Classification

Deep learning model for automated classification of cardiac arrhythmias using ECG signals from the MIT-BIH database. The project combines signal preprocessing via wavelet transform and a multi-layer CNN architecture, achieving over 98% test accuracy across 15 heartbeat classes. Designed for real-time and clinical applications.

Deterministic four-record MIT-BIH R-peak detection prestudy with traceable metrics, error analysis, tests, and provenance.

This project focuses on building an end-to-end ECG signal-processing and analysis pipeline using classical digital signal-processing techniques.

Reproducible pipeline for silent-failure auditing in ECG accept-sets (MIT-BIH) with Newton–Puiseux onset scoring

AI-powered ECG signal analysis framework built using Python and the MIT-BIH Arrhythmia Database. Features ECG visualization, R-peak detection, heart-rate estimation, and cardiac abnormality identification using expert cardiologist annotations.