MediScript is an experimental AI-driven healthcare platform designed to explore non-invasive blood group prediction using fingerprint images. The system integrates a deep learning model with a full-stack web application to demonstrate how biometric data and artificial intelligence can be combined to assist basic medical diagnostics.
The project was developed as part of an Artificial Intelligence and Machine Learning internship and serves as a proof-of-concept for applying computer vision techniques to healthcare applications. The platform allows users to upload fingerprint images and obtain predicted blood group classifications through an integrated machine learning model.
Traditional blood group identification methods require invasive blood sampling and laboratory testing. While these methods are reliable, they require specialized infrastructure and can be time-consuming. MediScript explores an alternative experimental approach that utilizes fingerprint images and deep learning models to predict blood groups.
The system uses a Convolutional Neural Network (CNN) implemented in PyTorch to analyze fingerprint ridge patterns and classify them into one of eight blood groups. The application integrates this prediction capability with a modern web interface built using the MERN stack.
The overall objective of the project is to demonstrate how artificial intelligence can be integrated with biometric analysis to create accessible and efficient healthcare tools.
According to the project report, the trained CNN model achieved approximately 92% training accuracy and around 88% validation accuracy during experimentation, demonstrating promising results for a prototype research system.
The MediScript system consists of three major components:
AI/ML Model
- A Convolutional Neural Network trained on grayscale fingerprint images.
- Responsible for extracting fingerprint features and predicting blood group classes.
Backend Services
- Node.js and Express-based REST APIs.
- Handles user authentication, data management, and prediction requests.
- Communicates with the AI model through a Python inference service.
Frontend Application
- React-based user interface.
- Provides features such as fingerprint upload, prediction results, doctor browsing, and appointment scheduling.
The complete system demonstrates the integration of machine learning pipelines with modern full-stack web development.
The core of MediScript is a CNN model implemented using PyTorch.
Model characteristics:
Input Grayscale fingerprint images resized to 256×256 pixels.
Architecture
- Convolution Layer (16 filters)
- Convolution Layer (32 filters)
- Max Pooling Layers
- Fully Connected Layer (128 neurons)
- Output Layer (8 blood group classes)
Output Classes
A+ A- B+ B- AB+ AB- O+ O-
Training configuration:
Loss Function CrossEntropyLoss
Optimizer Adam Optimizer
Batch Size 16
Epochs 5 (prototype training configuration)
The trained model is stored as a .pth file and loaded during runtime for inference.
AI-Based Blood Group Prediction Users can upload fingerprint images and obtain predicted blood group classifications.
Doctor Directory Displays available doctors along with specialization and availability information.
Appointment Booking Allows users to schedule appointments with listed doctors.
User Authentication Secure login and signup system using JWT tokens.
Profile Management Users can update their personal profile details.
Responsive Web Interface Built using React and Tailwind CSS for compatibility across desktop and mobile devices.
- User uploads a fingerprint image through the web interface.
- The image is sent to the backend server through an API request.
- The backend forwards the image to the Python-based AI inference service.
- The CNN model processes the fingerprint image and predicts the blood group.
- The predicted result is returned to the frontend and displayed to the user.
MediScript
│
├── backend
│ ├── controllers
│ ├── models
│ ├── routes
│ ├── middleware
│ └── server.js
│
├── frontend
│ ├── src
│ ├── components
│ ├── pages
│ └── App.jsx
│
├── assets
│
├── model.py
├── app.py
├── model.pth
└── README.md
Clone the repository
git clone https://github.com/devsudhanshu-bel/Mediscript.git
Navigate to the project directory
cd Mediscript
Install backend dependencies
cd backend
npm install
Install frontend dependencies
cd ../frontend
npm install
Run the backend server
npm start
Run the frontend development server
npm run dev
Run the AI prediction service
python app.py
This system is an experimental research prototype and is not intended for clinical or medical deployment.
The model was trained on a limited dataset of fingerprint images, which may affect prediction reliability for certain blood groups. Large-scale datasets and clinical validation would be required before such a system could be used in real-world healthcare environments.
Additionally, the current implementation relies on uploaded fingerprint images rather than real-time biometric scanner integration.
Potential improvements include:
- Training the model on larger and more diverse datasets
- Integrating real-time fingerprint scanner hardware
- Using advanced architectures such as ResNet or EfficientNet
- Deploying the system on scalable cloud infrastructure
- Developing a mobile application interface
- Integrating with hospital management systems
Sudhanshu Sreedhara Belavarthy, B.Tech Computer Science and Engineering (Artificial Intelligence and Machine Learning) Christ University, Bengaluru
This project is intended for academic and research purposes.