ECG Arrhythmia Detection

A CNN-based project for classifying ECG heartbeats using the MIT-BIH Arrhythmia dataset.

Project Overview

• Dataset: MIT-BIH Arrhythmia
• Input: 187-sample ECG signals per heartbeat
• Model: Convolutional Neural Network (CNN)
• Output: Predicted heartbeat class

Files

• arrhythmia.ipynb : Jupyter notebook with data preprocessing, model training, and evaluation.
• ecg_model.keras : Trained Keras model (can be converted to TensorFlow.js for web integration)

Notes

• This is a work-in-progress project for research and learning purposes.
• Demonstrates the full pipeline from data preprocessing to CNN-based ECG classification.

• How to Test Your ECG Data

You can input your own ECG signals to test the model using a CSV file.

1. CSV Format

• Each row should contain 187 numeric ECG values corresponding to a single heartbeat.

• Example (sample_ecg.csv):

• Each row represents one heartbeat.

2. Using the Web App / Backend

1. Go to the ECG prediction endpoint: /predict
2. Upload your CSV file containing ECG rows.
3. The model will return a predicted class for each heartbeat.

3. Sample Code (Python)

import requests

url = "http://localhost:5000/predict"
files = {'ecgfile': open('sample_ecg.csv', 'rb')}

response = requests.post(url, files=files)
print(response.json())



## Author
Shubham Thakur