Pneumonia Prediction from Chest X-Ray Images

A Flask and PyTorch web application that predicts whether a chest X-ray image is classified as NORMAL or PNEUMONIA. The project includes a trained convolutional neural network model, a browser-based upload interface, and a Jupyter notebook used for model development.

Important: This project is for educational and research purposes only. It is not a medical device and should not be used as a substitute for professional clinical diagnosis.

Features

• Chest X-ray image upload through a web dashboard
• Pneumonia vs normal classification
• Confidence score for the predicted class
• Separate probability scores for NORMAL and PNEUMONIA
• Custom CNN architecture built with PyTorch
• Image preprocessing with OpenCV and Torchvision transforms
• Flask API endpoint for inference
• Included notebook for model training and experimentation

Tech Stack

Area	Tools
Backend	Flask
Deep Learning	PyTorch, Torchvision
Image Processing	OpenCV, NumPy
Frontend	HTML, CSS, JavaScript, jQuery
Model Development	Jupyter Notebook

Project Structure

pneumonia_prediction/
|-- app.py
|-- best_pneumonia_model.pth
|-- pneumonia_model_creation.ipynb
|-- templates/
|   `-- dashboard.html
|-- Test sample images/
`-- README.md

Model Overview

The application uses a custom convolutional neural network with:

• 4 convolutional layers
• Max pooling after each convolution block
• Dropout for regularization
• Fully connected classification layers
• 2 output classes: NORMAL and PNEUMONIA

Input images are resized to 224x224, converted to tensors, and normalized before being passed to the model.

Getting Started

Prerequisites

• Python 3.8 or newer
• pip
• A trained model file named best_pneumonia_model.pth

Installation

Clone the repository:

git clone https://github.com/your-username/pneumonia_prediction.git
cd pneumonia_prediction

Create and activate a virtual environment:

python -m venv .venv
.venv\Scripts\activate

Install dependencies:

pip install flask torch torchvision opencv-python numpy

Run the App

python app.py

Open the application in your browser:

http://127.0.0.1:8080

Upload a chest X-ray image in PNG or JPEG format to generate a prediction.

API Usage

The app exposes a prediction endpoint:

POST /predict

Form field:

image: chest X-ray image file

Example JSON response:

{
  "prediction": "PNEUMONIA",
  "confidence": 96.42,
  "prob_normal": 3.58,
  "prob_pneumonia": 96.42
}

Dataset

The model notebook is designed around a chest X-ray pneumonia dataset with separate training, validation, and test folders.

Example dataset source:

https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia

Expected folder structure:

chest_xray/
|-- train/
|   |-- NORMAL/
|   `-- PNEUMONIA/
|-- val/
|   |-- NORMAL/
|   `-- PNEUMONIA/
`-- test/
    |-- NORMAL/
    `-- PNEUMONIA/

Training

Open the notebook to review or retrain the model:

jupyter notebook pneumonia_model_creation.ipynb

After training, save the best model weights as:

best_pneumonia_model.pth

The Flask app loads this file automatically on startup.