Our model is the most advanced ever made for gene assessment.
Precision Segmentation
Utilizes advanced instance segmentation models (e.g., Detectron2, Cascade Mask R-CNN, DeepLabV3+) to accurately isolate the mandibular condyle from MRI scans.
AI-Powered Diagnosis Support
Assists radiologists and researchers in identifying potential TMJ disorders with automated, consistent, and reproducible segmentation results.
Ensemble Learning Integration
Combines multiple model predictions using majority voting to enhance robustness and reduce false positives/negatives.
Augmented Dataset
Enhanced with data augmentation techniques (flip, rotate, scale, etc.) to improve model generalization across diverse MRI inputs.
Flexible Deployment
Supports both cloud-based (Google Colab) and local (GPU-enabled systems) environments for model training and evaluation.
Structured Dataset Format
Organized in COCO JSON format with clean folder structure, enabling seamless integration with computer vision frameworks.
Research-Oriented Design
Built with research in mind, enabling reproducibility, experimentation, and easy extension to other anatomical structures or segmentation tasks.
Acknowledgment
Thanks to the International Islamic University Malaysia (IIUM) for the collaboration with YARSI University in providing MRI data, as well as ethical support through IREC Approval No: IREC 2022-050. This greeting is also given to Chandra Prasetyo Utomo, M.S. as the supervising lecturer for his guidance and direction during this research process.
Sincere gratitude is also expressed to the YARSI University Artificial Intelligence Laboratory team, as well as all colleagues who have provided support, both technically and academically, in carrying out this research.
