OPG-SHAP: A Dental AI Tool for Explaining Learned Orthopantomogram Image Recognition
Issued Date
2024-01-01
Resource Type
Scopus ID
2-s2.0-85203604359
Journal Title
2024 International Technical Conference on Circuits/Systems, Computers, and Communications, ITC-CSCC 2024
Rights Holder(s)
SCOPUS
Bibliographic Citation
2024 International Technical Conference on Circuits/Systems, Computers, and Communications, ITC-CSCC 2024 (2024)
Suggested Citation
Hirunchavarod N., Dangsungnoen L., Thongprasant K., Phuphatham P., Prathansap N., Sributsayakarn N., Pornprasertsuk-Damrongsri S., Jirarattanasopha V., Intharah T. OPG-SHAP: A Dental AI Tool for Explaining Learned Orthopantomogram Image Recognition. 2024 International Technical Conference on Circuits/Systems, Computers, and Communications, ITC-CSCC 2024 (2024). doi:10.1109/ITC-CSCC62988.2024.10628217 Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/101229
Title
OPG-SHAP: A Dental AI Tool for Explaining Learned Orthopantomogram Image Recognition
Corresponding Author(s)
Other Contributor(s)
Abstract
We propose OPG-SHAP, a tool for extracting insight of the trained deep learning model for an Orthopantomograms (OPGs). The tool allows user to understand the areas that influence the model's predictions through SHAP (SHapley Additive exPlanations) interpretability. The tool automatically identify the regions within the OPG image that influence the model's decision-making through YOLO-based object detection model that trained to detect over 20 classes of dental region in OPGs with an AverageIoU = 0.795, mAP = 0.989, Precision = 0.986, Recall = 0.987, and F1 score = 0.986. When a user inputs an OPG image and model into the OPG-SHAP tool, the output highlights the positions the model considers important for its decision, indicated by positive and negative SHAP values.