🩺 Chest X-Ray Pneumonia Detection using Deep Learning

A deep learning project that classifies chest X-ray images into Normal and Pneumonia using PyTorch and Transfer Learning.

📑 Table of Contents

• Project Overview
• Features
• Dataset
• Technologies Used
• Model Architecture
• Results
• Installation
• Project Structure
• Future Improvements
• Author
• License

📌 Project Overview

Pneumonia is a serious lung infection that can be detected through chest X-ray images. Manual diagnosis can be time-consuming and depends on the expertise of radiologists.

This project uses a deep learning model based on PyTorch and Transfer Learning to automatically classify chest X-ray images into:

• Normal
• Pneumonia

The objective is to demonstrate how convolutional neural networks can assist in medical image classification.

✨ Features

• Data preprocessing and augmentation
• Transfer Learning using ResNet50
• Binary classification (Normal vs Pneumonia)
• Model training and validation
• Performance evaluation
• Prediction on new chest X-ray images
• Explainable AI with Grad-CAM (heatmaps of the regions influencing each prediction)

🔍 Explainability (Grad-CAM)

Grad-CAM produces a heatmap over the last convolutional layer showing which regions of the X-ray drove the prediction — useful for sanity-checking that the model looks at the lungs rather than artifacts.

Run it on any image from the command line:

python scr/gradcam.py --image path/to/xray.jpeg --model best_model.pth --output cam.png

Or use it from Python:

from scr.gradcam import load_model, run_gradcam

model = load_model("best_model.pth")
overlay, pred_idx, probs, cam = run_gradcam(model, "xray.jpeg")

A ready-to-run Grad-CAM cell is also included at the end of train_model.ipynb.

📂 Dataset

The dataset consists of chest X-ray images categorized into:

• Normal
• Pneumonia

Dataset Source

Google Drive:

https://drive.google.com/file/d/1d6G2RkXVNnMBJTO-QWCx6QZ2bMElMqgu/view

After downloading, extract the dataset into the dataset/ folder.

🛠 Technologies Used

• Python
• PyTorch
• Torchvision
• NumPy
• Matplotlib
• Pillow
• Scikit-learn

🧠 Model Architecture

This project uses a custom Convolutional Neural Network (CNN) designed for chest X-ray image classification.

Architecture:

Input Image (1 × 224 × 224) ↓ Conv2D (16 filters) + BatchNorm + ReLU + MaxPool ↓ Conv2D (32 filters) + BatchNorm + ReLU + MaxPool ↓ Conv2D (64 filters) + BatchNorm + ReLU + MaxPool ↓ Conv2D (128 filters) + BatchNorm + ReLU + MaxPool ↓ Conv2D (256 filters) + BatchNorm + ReLU + MaxPool ↓ Flatten ↓ Fully Connected (1024) ↓ Fully Connected (256) ↓ Fully Connected (128) ↓ Output Layer (2 classes: NORMAL / PNEUMONIA)

📈 Results

The trained model successfully classifies chest X-ray images into Normal and Pneumonia categories.

Evaluation metrics and visualizations will be added in future updates.

🚀 Installation

Clone the repository

git clone https://github.com/badivana/Chest-X-Ray-Classification.git

cd Chest-X-Ray-Classification

Install dependencies

pip install -r requirements.txt

📁 Project Structure

Chest-X-Ray-Classification/

├── README.md

├── requirements.txt

├── LICENSE

├── train_model.ipynb

├── best_model.pth

├── dataset/

├── images/

└── src/

🚀 Future Improvements

• Deploy using Streamlit
• Multi-class disease classification
• Vision Transformer implementation
• ✅ Explainable AI using Grad-CAM
• Docker support

👨‍💻 Author

Prajwal B T

Information Science & Engineering

NMAM Institute of Technology

GitHub: https://github.com/badivana

📄 License

This project is licensed under the MIT License.