PCOS Detection Tool

PCOS Detection Tool is a deep learning-based application designed to detect Polycystic Ovary Syndrome (PCOS) from medical images. This lightweight tool includes a trained deep learning model, a Flask backend for API-based predictions, and a simple HTML interface for user interaction.

Table of Contents

• Overview
• Installation
• Dataset Requirements
• Usage
  • Training the Model
  • Running the Flask App
• File Structure
• Flask API
• Frontend
• Dependencies
• Contributing
• License
• Contact

Overview

This repository contains the code and assets required for the PCOS Detection Tool, which includes:

• Training Model: A CNN built with TensorFlow/Keras for image classification.
• Flask Backend (app.py): Loads the trained model and provides an API endpoint for image predictions.
• HTML Frontend (index.html): A user-friendly interface that allows users to upload images and view classification results.
• Dataset: A dataset of medical images organized into class-specific directories.

Installation

1. Clone the repository:

   git clone https://github.com/your_username/pcos-detection-tool.git
   cd pcos-detection-tool

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate  # For Linux/MacOS
   venv\Scripts\activate     # For Windows

3. Install dependencies:

   pip install -r requirements.txt

   Example requirements.txt:

   tensorflow
   keras
   flask
   pillow
   numpy
   opencv-python
   

Dataset Requirements

To train or retrain the model, you must have a properly structured dataset of medical images:

• Directory Structure:
  Create a main dataset folder (e.g., dataset_extracted/PCOS) with two subdirectories:

  • Infected — Contains images showing signs of PCOS.
  • Not Infected — Contains images without PCOS.

• Image Formats:
  Supported image formats include JPEG, PNG, and JPG.

• Preprocessing:
  Images will be automatically resized to 128x128 pixels and normalized to have pixel values between 0 and 1.

Ensure your dataset follows this structure before running the training script.

Usage

Training the Model

If you plan to train the model on your own dataset:

1. Organize your dataset as described above.

2. Run the training script (e.g., your_model_script.py):

   python your_model_script.py

This script will:

• Load and preprocess images.
• Split data into training and testing sets.
• Train the CNN using early stopping and model checkpoint callbacks.
• Generate evaluation plots (accuracy, loss, ROC, and Precision-Recall curves).
• Save the trained model as model.h5 (or similar).

Running the Flask App

Once the model is trained (or if you have a pre-trained model):

1. Ensure the model file (e.g., model.h5) is in the project directory.

2. Start the Flask application:

   python app.py

3. Open your browser and navigate to http://127.0.0.1:5000/ to access the tool.

Uploading an Image

• Use the web interface provided in index.html to upload an image.
• Click the Predict button to receive the classification result (either "Infected" or "Not Infected").

File Structure

pcos-detection-tool/
├── app.py               # Flask backend application
├── your_model_script.py # Script for training and evaluating the CNN model
├── model.h5             # Trained deep learning model
├── templates/
│   └── index.html       # Frontend user interface
├── requirements.txt     # Python dependencies
├── README.md            # Project documentation (this file)

Flask API

• Endpoint: POST /predict
• Input: An image file (JPEG/PNG)
• Output: JSON response with a classification result ("Infected" or "Not Infected")

Frontend

The HTML interface (index.html) provides:

• Image Upload: A drag-and-drop/file selection area with image preview.
• Result Display: Shows the prediction result dynamically.
• Dark Mode Toggle: Allows switching between light and dark themes.

Dependencies

This project uses the following libraries:

• TensorFlow/Keras: For building and running the deep learning model.
• Flask: For creating the backend API.
• OpenCV and Pillow: For image processing.
• NumPy: For numerical computations.

Contributing

Contributions are welcome! Feel free to:

• Report bugs or request features via the issue tracker.
• Fork the repository and submit pull requests for improvements.
• Update documentation and code for clarity.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For questions or suggestions, please get in touch with me, Anish Sharma at sharmaanish310@gmail.com.