MediScan AI

A comprehensive AI-powered medical diagnostic platform that leverages state-of-the-art machine learning models to assist in the early detection and analysis of various medical conditions through image and data analysis.

Overview

MediScan AI is an innovative web-based application that integrates multiple deep learning and machine learning models to provide diagnostic assistance for seven major medical conditions:

• 🦠 COVID-19 Detection - Chest X-ray analysis for pneumonia patterns
• 🧠 Brain Tumor Detection - MRI scan analysis for tumor identification
• 👩 Breast Cancer Detection - Predictive analysis using clinical parameters
• 🧠 Alzheimer's Disease Detection - MRI-based dementia classification
• 💉 Diabetes Prediction - Risk assessment using physiological parameters
• 🫁 Pneumonia Detection - Chest X-ray analysis for lung infections
• ❤️ Heart Disease Prediction - Cardiovascular risk assessment

Demo

Video Demo (Watch this) - https://drive.google.com/file/d/1v5stMRkBzpqc0mnqDkDXjfaOMKHAF4El/view?usp=sharing

✨ Key Features

Advanced AI Diagnostics

• Multi-Modal Analysis: Supports both image-based (X-rays, MRIs) and parameter-based diagnostics
• Real-Time Predictions: Instant results with confidence scores
• User-Friendly Interface: Intuitive web interface accessible to healthcare professionals and researchers

Comprehensive History Tracking

• Detection History: Complete log of all diagnostic sessions
• Data Persistence: CSV-based storage for analysis and auditing
• Export Capabilities: Easy data export for further research

Robust Architecture

• Modular Design: Clean separation of models and web interface
• Error Handling: Graceful degradation when models are unavailable
• Scalable Framework: Easy to extend with new diagnostic models

Web Interface

• Responsive Design: Works seamlessly across devices
• Bootstrap Framework: Modern, professional UI
• Interactive Forms: Guided input collection for accurate diagnostics

🛠️ Technology Stack

Backend Framework

• Flask 3.0.0: Lightweight Python web framework
• Python 3.9.13: Stable runtime environment

Machine Learning & AI

• TensorFlow 2.12.0: Deep learning framework for CNN models
• Keras: High-level neural network API
• Scikit-learn 0.24.2: Traditional ML algorithms
• XGBoost 2.0.3: Gradient boosting for tabular data

Computer Vision

• OpenCV 4.5.1.48: Image processing and computer vision
• Imutils 0.5.4: Image processing utilities

Data Processing

• NumPy: Numerical computing
• Pandas: Data manipulation (implied through sklearn)
• Joblib: Model serialization

Machine Learning Models

Deep Learning Models (CNN-based)

COVID-19 Detection Model

• Architecture: Convolutional Neural Network with 3 Conv2D layers
• Input: Chest X-ray images (224x224x3)
• Output: Binary classification (COVID/Normal)
• Training Data: ~5,000 chest X-ray images
• Accuracy: ~95% on validation set
• Technique: Transfer learning with custom CNN layers

Brain Tumor Detection Model

• Architecture: VGG16-based CNN with custom preprocessing
• Input: Brain MRI scans (224x224x3)
• Output: Binary classification (Tumor/No Tumor)
• Preprocessing: Advanced cropping and skull removal
• Technique: Image segmentation and classification

Alzheimer's Disease Model

• Architecture: Custom CNN with 4 convolutional blocks
• Input: Brain MRI slices (176x176x3)
• Output: Multi-class classification (4 dementia stages)
• Classes: NonDemented, VeryMildDemented, MildDemented, ModerateDemented
• Technique: Multi-stage classification with batch normalization

Pneumonia Detection Model

• Architecture: Sequential CNN with dropout regularization
• Input: Chest X-ray images (150x150x3)
• Output: Binary classification (Pneumonia/Normal)
• Training Data: Large chest X-ray dataset
• Technique: Data augmentation and regularization

Traditional Machine Learning Models

Breast Cancer Prediction

• Algorithm: XGBoost Classifier
• Features: 5 clinical parameters (concave points, area, radius, perimeter, concavity)
• Output: Binary classification (Malignant/Benign)
• Technique: Ensemble learning with gradient boosting

Diabetes Prediction

• Algorithm: Random Forest Classifier (saved as pickle)
• Features: 8 physiological parameters
• Output: Binary classification (Diabetic/Non-Diabetic)
• Technique: Ensemble of decision trees

Heart Disease Prediction

• Algorithm: Custom ML model (pickle format)
• Features: Clinical parameters (age, blood pressure, cholesterol, etc.)
• Output: Binary classification (Disease/No Disease)
• Technique: Traditional supervised learning

System Architecture

MediScan AI/
├── app.py                 # Flask web application
├── models/                # Pre-trained ML models
│   ├── covid.h5          # COVID detection model
│   ├── braintumor.h5     # Brain tumor model
│   ├── alzheimer_model.h5 # Alzheimer's model
│   ├── pneumonia_model.h5 # Pneumonia model
│   ├── cancer_model.pkl  # Breast cancer model
│   ├── diabetes.sav      # Diabetes model
│   └── heart_disease.pickle.dat # Heart disease model
├── templates/            # HTML templates
│   ├── homepage.html
│   ├── covid.html
│   ├── resultc.html
│   └── ...
├── static/               # Static assets
│   ├── uploads/         # User uploaded images
│   └── images/          # UI images
├── detection_history.csv # Diagnostic history log
└── tools/               # Utility scripts

🚀 Getting Started

Prerequisites

• Python 3.9.13
• Conda package manager
• Git

Installation

1. Clone the repository

git clone <repository-url>
cd MediScan-AI

2. Create Conda Environment

conda create -n mediscan python=3.9.13
conda activate mediscan

3. Install Dependencies

pip install opencv-python==4.5.1.48 numpy tensorflow==2.12.0 scikit-learn==0.24.2 imutils==0.5.4 flask==3.0.0 xgboost==2.0.3

Running the Application

1. Start Flask Server

flask run

2. Access the Application Open your browser and navigate to: http://127.0.0.1:5000

Usage Guide

For Image-Based Diagnostics

1. Navigate to the desired diagnostic page (COVID, Brain Tumor, etc.)
2. Fill in patient information (name, age, gender, etc.)
3. Upload the medical image (X-ray, MRI, etc.)
4. Click "Submit" for instant AI-powered analysis
5. View results with confidence scores

For Parameter-Based Diagnostics

1. Select the appropriate diagnostic module
2. Enter clinical parameters in the form
3. Submit for risk assessment
4. Review prediction results

Viewing History

• Access the History page to view all previous diagnostics
• Export data for further analysis
• Clear history when needed

🔍 Model Performance

Model	Type	Accuracy	Input Type	Output Classes
COVID-19	CNN	~95%	Chest X-ray	2 (COVID/Normal)
Brain Tumor	CNN	~92%	Brain MRI	2 (Tumor/No Tumor)
Alzheimer's	CNN	~88%	Brain MRI	4 (Dementia Stages)
Pneumonia	CNN	~94%	Chest X-ray	2 (Pneumonia/Normal)
Breast Cancer	XGBoost	~96%	Clinical Data	2 (Malignant/Benign)
Diabetes	Random Forest	~85%	Physiological Data	2 (Diabetic/Normal)
Heart Disease	ML Model	~83%	Clinical Data	2 (Disease/No Disease)

Advanced Configuration

Model Loading

The application automatically loads all models on startup. If a model fails to load, the corresponding diagnostic feature becomes unavailable with appropriate user messaging.

Image Preprocessing

• Standardization: All images normalized to [0,1] range
• Resizing: Consistent input dimensions for each model
• Augmentation: Training data enhanced with rotations, flips, and zooms

Error Handling

• Model Unavailable: Graceful fallback with user notification
• Invalid Input: Form validation and error messages
• File Upload: Secure file handling with type validation

Future Enhancements

Planned Features

• Model Explainability: Integration of SHAP/LIME for prediction explanations
• Multi-Modal Fusion: Combining multiple imaging modalities
• Real-Time Training: Online learning capabilities
• API Endpoints: RESTful API for integration
• Batch Processing: Multiple image analysis
• Performance Metrics: Detailed accuracy and ROC curves

Research Directions

• Federated Learning: Privacy-preserving model training
• Transfer Learning: Cross-domain model adaptation
• Uncertainty Quantification: Confidence intervals for predictions
• Clinical Validation: Real-world performance studies

🤝 Contributing

We welcome contributions! Please follow these steps:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Development Guidelines

• Follow PEP 8 style guidelines
• Add docstrings to new functions
• Update tests for new features
• Ensure models are version-controlled appropriately

License

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

Disclaimer

MediScan AI is a research and educational tool, not a clinical diagnostic device.

• Models are trained on publicly available datasets
• Results should not be used for actual medical decisions
• Always consult qualified healthcare professionals
• Performance may vary with real-world data
• Regular model updates and validation recommended

Support

For questions, issues, or contributions:

• Open an issue on GitHub
• Contact the development team
• Check the documentation for common solutions

Acknowledgments

• Medical imaging datasets from various public repositories
• Open-source ML frameworks and libraries
• Research community for published methodologies
• Healthcare professionals for domain expertise

---

MediScan AI - Advancing healthcare through artificial intelligence 🚀