AI-Based Multi-Stage Pipeline for Detecting Caries and Restorations in Pediatric Primary Molars from Intraoral Photographs
Description
This repository contains the computational and documentation materials associated with the study “AI-Based Multi-Stage Pipeline for Detecting Caries and Restorations in Pediatric Primary Molars from Intraoral Photographs.” The repository currently includes the following materials: * `pipeline.ipynb`: Jupyter Notebook containing the main deep learning pipeline and computational workflow. * `interface script.ipynb`: Jupyter Notebook containing the inference/interface implementation for applying the trained pipeline. * `Dental_YOLO_SAM_CNN_Documentation.docx`: Technical documentation describing the computational workflow and implementation. * `runs/`: Model training and experiment outputs. * `test_image.JPG` and `test_image2.JPG`: Example intraoral images provided for testing and demonstration of the inference workflow. The study uses a multi-stage deep learning pipeline for automated detection, segmentation, and health-status classification of primary molars in intraoral photographs. The computational workflow includes YOLOv8 for tooth detection, the Segment Anything Model (SAM) for tooth segmentation, and ResNet50 for five-class health-status classification. The repository is provided to support computational transparency, reproducibility, and independent inspection of the implemented workflow. No patient-identifiable information is included in the repository.
Files
Institutions
- Shahid Beheshti University of Medical SciencesTehran, Tehran