InVisionDX

A comprehensive Flask-based web application for automated medical diagnosis through image analysis. This platform leverages machine learning models to predict various medical conditions, all while offering a secure and modern user experience.

Access the ML models here--> https://huggingface.co/shrrynsh

Index

• Project Overview
• Key Features
  • User Authentication System
  • Medical Image Analysis
  • Machine Learning Integration
  • Database Management
  • Frontend Features
  • Technical Stack
  • Security Features
• Notable Implementation Details
  • Image Processing Pipeline
  • Prediction System
  • User Experience
  • Data Management
• Project Structure
• Installation
• Usage

Project Overview

This web application is designed to analyze medical images and predict conditions such as COVID-19, Pneumonia, Tuberculosis, Lung Cancer, and Alzheimer's disease. The project integrates multiple machine learning models, robust user authentication, and a responsive web interface to provide an end-to-end diagnostic solution.

Key Features

User Authentication System

• Registration & Login: Secure user sign-up and sign-in.
• Password Security: Password hashing and recovery functionality.
• Session Management: Maintain user sessions for seamless navigation.
• User Profiles: Dashboard displaying user details and prediction history.

Medical Image Analysis

• Multi-Disease Prediction:
  • COVID-19 detection
  • Pneumonia detection
  • Tuberculosis (TB) detection
  • Lung Cancer detection
  • Alzheimer's disease detection
• Image Preprocessing: Each prediction model has its own preprocessing routine to resize, normalize, and support various image formats.

Machine Learning Integration

• TensorFlow Models:
  • COVID_Detect.keras
  • pneumonia_predict.keras
  • tb_model_final.keras
  • best_model_lung_cancer.keras
  • alzheimer_detect.keras
• On-Demand Loading: Models are loaded as needed, optimizing memory usage.
• Prediction Outputs: Provides confidence scores for each prediction, with results stored in the database.

Database Management

• SQLite & SQLAlchemy: Manages user data, login activities, and historical prediction records.
• Data Persistence: Tracks user interactions and stores prediction history for future reference.

Frontend Features

• Responsive Design: Modern UI that works seamlessly across devices.
• Interactive Forms: Easy image uploads and real-time feedback.
• Dashboard & Profile Management: View and manage prediction history and personal information.
• Additional Pages: Informative About page detailing the project.

Technical Stack

• Backend: Flask 2.0.1
• Database: SQLite managed with SQLAlchemy
• Authentication: Flask-Login for secure access control
• Machine Learning: TensorFlow 2.8.0
• Image Processing: OpenCV and Pillow
• Frontend: HTML, CSS, JavaScript
• Utilities: NumPy, email-validator, and others

Security Features

• Password Hashing & CSRF Protection: Ensuring secure user interactions.
• Session Management & Protected Routes: Prevent unauthorized access.
• Secure File Handling: Safeguarding image uploads and model files.

Notable Implementation Details

Image Processing Pipeline

• Custom Preprocessing: Unique functions for each medical condition ensure images meet model requirements.
• Format Support: Handles various image formats, resizing and normalization included.

Prediction System

• On-Demand Model Loading: Efficient memory use by loading models as needed.
• Result Storage: Prediction outcomes, including confidence scores, are saved to the database.
• User History: Enables users to track and review their past predictions.

User Experience

• Interactive and Validated Forms: Improves data input reliability and user satisfaction.
• Responsive Design & Animations: Utilizes modern UI elements and animations (powered by vanta.js) for an engaging experience.
• Error Handling: Provides clear feedback in case of issues.

Data Management

• User Profiles: Manage personal data and view prediction history.
• Activity Tracking: Logs user login times and interactions.

Project Structure

plaintext mywebsite/ ├── app.py # Main application file ├── requirements.txt # Project dependencies ├── templates/ # HTML templates for the web pages ├── static/ # Static assets (CSS, JavaScript, images) ├── instance/ # Instance-specific files and configurations └── *.keras # Pre-trained machine learning model files

Installation

• Follow these steps to set up the project locally:

  1. Clone the Repository: bash git clone https://github.com/shrrynsh/InVisionDX.git cd yourrepository

  2. Create a virtual environment: bash python3 -m venv venv source venv/bin/activate

  3. Install Dependencies: bash pip install -r requirements.txt

  4. Make sure that the file paths of models in app.py are correct.

Usage

• Start the Application: bash python3 app.py

• Access the Application: Open your browser and navigate to http://127.0.0.1:5500.

• Register/Login: Create a new account or log in with existing credentials. Use your dashboard to upload images and view predictions.

• Making Predictions: Upload a medical image using the interactive form. View the prediction results along with confidence scores. Check your prediction history in your user profile.