Breast Cancer Detection Demo API

Description

This project is a web application demonstrating binary classification (benign vs. malignant) of breast cancer histopathology images. It uses a fine-tuned MobileNetV3 Large PyTorch model served via a Flask API, featuring a simple web interface for interactive predictions.

Current Time: Sunday, March 30, 2025 at 11:36:04 PM Location: Charlotte, North Carolina, United States

(Disclaimer: This is currently a demonstration project and is NOT intended for actual medical diagnosis or clinical use.)

Current Features

• Web Interface (/): Allows users to upload an image file and view the model's prediction (Benign/Malignant).
• Prediction API (/predict - POST): Enables programmatic predictions by sending Base64 encoded image data within a JSON payload.
• Health Check (/health - GET): A simple endpoint to verify if the application is running.
• Model: Utilizes a MobileNetV3 Large model, pre-trained on ImageNet and fine-tuned for this classification task, loaded from a local checkpoint.

Tech Stack

• Backend: Python, Flask
• ML/Image Processing: PyTorch, TorchVision, Pillow (PIL)
• WSGI Server (Optional): Gunicorn
• Frontend: HTML, CSS (Bootstrap 5), JavaScript (Fetch API)

Setup & Installation

Prerequisites:

• Python 3.x
• pip (Python package installer)
• Git

Steps:

1. Clone the repository:

   git clone <your-repository-url>
   cd <your-repository-directory>

2. Create a virtual environment (Recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies: (Note: You should create a requirements.txt file first by running pip freeze > requirements.txt after installing necessary packages locally)

   pip install -r requirements.txt

   (Likely requirements include: flask, torch, torchvision, Pillow, gunicorn)

4. Obtain Model Checkpoint:

   • Create a directory named model in the project root.
   • Place your trained model checkpoint file (e.g., mobilenet_checkpoint.tar) inside the model/ directory.

5. Run the Application:

   • Development Server:

     python app.py

     Access the application at http://127.0.0.1:5000 (or the port specified in app.py).

   • Production Server (using Gunicorn):

     # Example: 4 workers, listening on all interfaces, port 8090
     gunicorn -w 4 -b 0.0.0.0:8090 app:app

     Access the application at http://<your_server_ip>:8090.

Usage

1. Web Interface:

• Navigate to the root URL (http://<host>:<port>/) in your web browser.
• Click the "Choose File" button to select an image.
• An image preview will be displayed.
• Click the "Predict" button.
• The prediction result ("Benign" or "Malignant") will appear below the button.

2. Health Check API:

• Send a GET request to the /health endpoint.

  curl http://<host>:<port>/health

• Expected response: {"status":"healthy"}

3. Prediction API:

• Send a POST request to the /predict endpoint with a JSON payload containing the Base64 encoded image string.

  # Example using curl:
  
  # 1. Base64 encode your image (example on Linux/macOS)
  # Replace 'path/to/your/image.png' with the actual image path
  IMAGE_B64=$(base64 path/to/your/image.png | tr -d '\n')
  
  # 2. Send the POST request (adjust URL and port as needed)
  curl -X POST \
    [http://127.0.0.1:5000/predict](http://127.0.0.1:5000/predict) \
    -H 'Content-Type: application/json' \
    -d '{
      "image": "'"${IMAGE_B64}"'"
    }'

• Expected response: {"prediction":"benign"} or {"prediction":"malignant"}

Project Structure

├── PyPI-Packages.md
├── README.md
├── __pycache__
│   ├── app.cpython-312.pyc
│   └── image_transformations.cpython-312.pyc
├── app.py
├── data
│   ├── README.md
│   ├── benign
│   ├── license.txt
│   └── malignant
├── endpoint_validation.py
├── image_transformations.py
├── model
│   └── mobilenet_checkpoint.tar
├── results
└── templates
    └── index.html

7 directories, 11 files

Examples / Screenshots

Here's how the web interface looks after uploading an image and receiving a prediction:

Benign Detection Example:

Malignant Detection Example:

---

Dockerfile

docker build -t breast-cancer-classifier:v1.0 . docker run -p 8090:8000 --name breast-cancer-classifier breast-cancer-classifier:v1.0 docker tag breast-cancer-classifier:v1.0 awesomeai/breast-cancer-classifier:v1.0