Heart Disease Prediction Using ECG and Vision Transformer (ViT)

This project combines machine learning and computer vision techniques to analyze electrocardiogram (ECG) data and classify heart conditions into four categories:

1. Myocardial Infarction
2. Abnormal Heartbeat
3. History of Myocardial Infarction
4. Normal Heartbeat

The project uses the Vision Transformer (ViT) model fine-tuned on ECG-XRAY images and is hosted on the Hugging Face Model Hub.

---

Repository Under: AcWoC'25

Club: Android Club, VIT Bhopal University

---

Table of Contents

1. Dataset
2. Features
3. Requirements
4. Usage
5. Training the Model
6. Evaluating the Model
7. Classifying User-Provided Images
8. Model Deployment
9. Results
10. References
11. Acknowledgments

---

Dataset

The dataset used for training and evaluation is located in the Dataset directory and can also be accessed from the Hugging Face Dataset Hub:
ECG-XRAY Dataset

It consists of ECG recordings labeled according to the aforementioned categories.

---

Features

• Vision Transformer (ViT): Fine-tuned for medical image classification.
• ECG Analysis: Classifies images into four heart condition categories.
• Evaluation Metrics: Includes accuracy, precision, recall, and confusion matrices.
• User Input Classification: Supports real-time ECG image classification.
• Model Hosting: Published on Hugging Face Hub for easy reuse.

---

Requirements

Ensure you have the necessary Python packages installed before running the project. Dependencies include:

• transformers
• datasets
• torch
• Pillow
• scikit-learn

Install them using:

pip install -r requirements.txt