AI-Powered Breast Cancer Diagnosis App by Hannah Pham

Breast Cancer Diagnosis Web Application

Overview

This project is a Flask-based web application that utilizes machine learning models to assist in the prediction of breast cancer risk based on user-provided mammogram data. It integrates multiple ML algorithms and provides a comparative analysis of predictions to aid in educational and research purposes.

Key Features

Interactive Web Interface:

A user-friendly frontend built with HTML, CSS, and Flask templates for seamless interaction.

Machine Learning Integration:

Supports predictions from four models:

K-Nearest Neighbors (KNN)
Support Vector Machine (SVM)
Decision Tree
Neural Network

Comparative Analysis:

Displays predictions from all models with probabilities and provides overall risk analysis using soft voting (average probabilities) and majority voting.

Data Security:

Processes user inputs locally during the session, ensuring privacy and data protection.

Deployment-Ready:

Includes a Dockerized application for deployment, with plans for AWS-based CI/CD pipeline integration.

Supports predictions from four models:

Programming Languages: Python
Libraries & Frameworks: Flask, Pandas, NumPy, Scikit-learn, TensorFlow/Keras
Frontend: HTML, CSS
Deployment Tools: Docker, AWS (planned for CI/CD)

Key Contributions

Designed and implemented the machine learning pipeline, including data preprocessing, model training, and evaluation.>
Developed a dynamic web interface allowing users to input mammogram data and visualize predictions.
Deployed the app using Docker and prepared for integration with AWS for scalable deployment.
Ensured robust comparative analysis by displaying results from multiple models and supporting majority voting and soft voting.

How to Use

Navigate to the home page and enter the mammogram info.
Click "Analyze Now" to submit the data.
View predictions from all four models and an overall diagnosis.
Review suggestions and next steps, emphasizing the importance of consulting a healthcare professional.

Outcomes

Demonstrated the power of machine learning in medical research by achieving 96% accuracy with various models.
Highlighted the importance of feature engineering and model evaluation techniques in healthcare applications.
Delivered a fully functional web application showcasing both technical and problem-solving skills in machine learning, software engineering, and deployment.

Next Steps

AWS Deployment: Implement CI/CD pipeline using AWS CodePipeline, CodeDeploy, and Elastic Beanstalk.
Explainable AI (XAI): Integrate SHAP or LIME to provide interpretability for predictions.
Model Optimization: Experiment with hyperparameter tuning and ensemble methods to further improve accuracy.