Degree

Bachelor of Science (Computer Science)

Department

Department of Computer Science

School

School of Mathematics and Computer Science (SMCS)

Advisor

Ms. Abeera Tariq, Lecturer, Department of Computer Science

Co-Advisor

Farrukh Saif - Salfsoft Technologies

Keywords

Dental X-ray Analysis, Healthcare Web Application, Medical Image Segmentation

Abstract

OrthoVision is a web-based AI-powered dental diagnostic and management system developed to address the critical gaps in dental care delivery in Pakistan. The platform leverages deep learning techniques to analyze periapical dental X-rays and detect conditions such as caries, crowns, periapical lesions, and root canals. By providing AI-generated diagnostic support, it aims to assist dentists with accurate, consistent, and efficient evaluations, reducing diagnostic errors and saving time. Beyond diagnosis, OrthoVision offers a centralized digital solution for managing hospital operations. It includes role-specific portals for patients, dentists, and administrators, enabling appointment scheduling, report access, and real-time staff coordination. This improves workflow, reduces administrative burden, and ensures better continuity of care across facilities. The project contributes to the modernization of Pakistan’s dental healthcare system by offering a cost-effective, accessible, and scalable platform. With a strong focus on usability, security, and future integration of features like multilingual voice assistance etc, OrthoVision aspires to become a transformative tool in both urban and underserved healthcare environments.

Tools and Technologies Used

Next.js 15.1.4 (React Framework) TypeScript Tailwind CSS Radix UI Components React 19 Axios for API calls Various UI libraries (lucide-react, react-day-picker, sonner) Backend: Django (backend framework) Django REST Framework PostgreSQL PyTorch TensorFlow Scikit-learn OpenCV Ultralytics YOLO Channels for WebSocket support Redis for real-time features Others: Git (version control) Kaggle (for GPU training), CloudFare, Docker, AWS

Methodology

Frontend Design The frontend was developed using Next.js (React framework) with TypeScript, Tailwind CSS, and React Query. The application’s structure was modular, with components, hooks, contexts, utility libraries, and public assets organized to support maintainability and scalability. The frontend’s primary responsibilities included rendering the user interface, making REST/GraphQL API calls, and receiving real-time updates via WebSocket for dynamic features such as appointment status and X-ray result notifications.

Backend Architecture The system followed a modular monolithic architecture, where key backend components were organized into distinct modules communicating through internal RESTful APIs. The core infrastructure included:

Core Service: Routed incoming requests, managed logging, configuration, and system-wide settings.

Authentication Service: Handled user login and authentication using JWT and Role-Based Access Control (RBAC) for secure access.

Clinic Service: Managed clinic data, doctor details, and scheduling.

Appointment Service: Responsible for booking, editing, and canceling appointments while handling conflict management.

X-ray Analysis Service: Managed image uploads, triggered AI model analysis, and stored the results.

Notification Service: Sent real-time notifications related to appointments and X-ray analysis via WebSocket.

The backend modules worked in tandem to provide a seamless user experience while ensuring separation of concerns and scalability.

AI Model Integration The machine learning pipeline involved fine-tuning various YOLO variants on the custom-labeled periapical dental X-ray dataset. This included both object detection and segmentation tasks. Data augmentation techniques were applied, and hyperparameter tuning was carried out to optimize model performance. After evaluating the models using mean Average Precision (mAP), YOLOv11s was selected for its superior segmentation accuracy and was fine-tuned for the final deployment and inference.

Database Layer The system utilized a PostgreSQL database to store essential data such as:

User Data

Appointment Information

X-ray Results

Clinic & Doctor Metadata

Each backend service interacted with the appropriate database models, ensuring consistent and efficient data handling. The database schema was designed to match the service responsibilities, allowing clear and reliable data access.

Security & Reliability The platform implemented comprehensive security measures, including:

JSON Web Tokens (JWT) for user authentication.

Role-Based Access Control (RBAC) for managing access based on user roles.

Input validation and rate limiting to prevent malicious activity and ensure system stability.

Transaction management to ensure data consistency and integrity across services.

This methodology ensured a robust, user-friendly, and scalable platform capable of integrating advanced machine learning techniques in a clinical setting, with an emphasis on security, real-time performance, and data reliability

Document Type

Restricted Access

Submission Type

BSCS Final Year Project

Download

Share

COinS