Scaneurysm AI - Brain CT Aneurysm Detection

An AI-powered system for detecting brain aneurysms in CT scan images using PyTorch and AWS SageMaker.

Overview

This project implements a deep learning model that analyzes brain CT scans to detect potential aneurysms. It uses a VGG16-based architecture and can be deployed as a serverless endpoint on AWS SageMaker.

Prerequisites

• Python 3.9+
• AWS Account with SageMaker access
• Docker (for local testing)
• Required Python packages (see requirements.txt)

Setup

1. Create and activate virtual environment:

./setup_venv.sh

2. Install dependencies:

pip install -r requirements.txt

3. Setup AWS infrastructure:

./setup_aws_infrastructure.sh

Project Structure

.
├── my_model/
│   ├── model.pth              # Your trained model weights
│   └── code/
│       ├── inference.py       # Model inference code
│       └── requirements.txt   # Inference dependencies
├── test-local.ipynb          # Local testing notebook
├── final.ipynb               # Deployment notebook
├── setup_aws_infrastructure.sh
├── setup_venv.sh
├── requirements.txt
└── Dockerfile

Model Details

• Architecture: Modified VGG16
• Input: Brain CT scan images (224x224 RGB)
• Output: Binary classification (Aneurysm/Non-aneurysm)
• Framework: PyTorch 2.1.0

Deployment

1. Package your trained model:

# Ensure your model.pth is in the my_model/ directory
tar -czf model.tar.gz my_model/

2. Open final.ipynb in Jupyter and follow the deployment steps.

3. The deployment will create a serverless SageMaker endpoint.

Testing

1. Local testing:

   • Use test-local.ipynb to verify model functionality

2. Endpoint testing:

   • Use the provided test functions in final.ipynb
   • Test with sample CT scan images via URLs

API Usage

import boto3
import json

def invoke_endpoint(endpoint_name, image_url):
    runtime = boto3.client('sagemaker-runtime')
    response = runtime.invoke_endpoint(
        EndpointName=endpoint_name,
        ContentType='application/json',
        Body=json.dumps({"url": image_url})
    )
    return json.loads(response['Body'].read().decode())

Environment Variables

Required AWS environment variables:

• AWS_ACCESS_KEY_ID
• AWS_SECRET_ACCESS_KEY
• AWS_DEFAULT_REGION

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a new Pull Request

License

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

Acknowledgments

• VGG16 architecture from torchvision
• AWS SageMaker team for serverless inference capabilities

Important Note

Remember to bring your own trained model (model.pth). This repository does not include pretrained weights.