COVID-19 and Pneumonia Chest X-ray Classification System

This project presents an AI-powered diagnostic system that classifies chest X-ray images as either Normal or Pneumonia, including cases caused by COVID-19. It utilizes a convolutional neural network (ResNet18), trained on a labeled dataset of X-ray images, and provides an interface for image upload and prediction through a Flask-based web API and a simple frontend.

Features

• Deep learning model using PyTorch and ResNet18
• Trained on chest X-ray images categorized as "Normal" or "Pneumonia"
• Flask backend for real-time predictions
• Frontend built with HTML, CSS, and JavaScript for file upload and result display
• Test accuracy exceeding 95% on clean datasets

Project Structure

covid-xray-ai/
├── app.py                # Flask backend application
├── train_model.py        # Script to train and save the model
├── model/
│   └── covid_model.pth   # Trained PyTorch model weights
├── dataset/
│   ├── train/
│   │   ├── NORMAL/
│   │   └── PNEUMONIA/
│   └── test/
│       ├── NORMAL/
│       └── PNEUMONIA/
├── static/
│   └── index.html        # Frontend UI for prediction
└── README.md

Requirements

• Python 3.8+
• PyTorch
• torchvision
• Flask
• PIL (Pillow)

To install dependencies:

pip install torch torchvision flask pillow

How to Use

1. Prepare the Dataset

Place your training and testing data in the following structure:

dataset/
├── train/
│   ├── NORMAL/
│   └── PNEUMONIA/
└── test/
    ├── NORMAL/
    └── PNEUMONIA/

Each folder should contain relevant chest X-ray image files.

2. Train the Model

Run the training script:

python train_model.py

After training, the model will be saved as model/covid_model.pth.

3. Start the Web Server

Run the Flask app:

python app.py

This will start the backend at http://localhost:5000.

4. Use the Frontend

Open static/index.html in your web browser. Upload an X-ray image, and the system will return a prediction: NORMAL or PNEUMONIA.

NORMAL

PNEUMONIA

Model Information

• Architecture: ResNet18
• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Input Size: 224x224 (resized)
• Number of Classes: 2 (NORMAL, PNEUMONIA)

Evaluation

The model achieved a test accuracy of 100% on the provided test dataset. Real-world performance may vary depending on image quality and dataset variability. Further validation on diverse datasets is recommended before clinical use.

License

This project is open source and may be used for educational and research purposes. It is not intended for medical diagnostics without proper validation and regulatory approval.