Petreska, Anita and Nikolovska, Julijana and Zabokova Bilbilova, Efka and Kokolanski, Vlatko and Ristevski, Blagoj and Rendevski, Nikola (2026) Explainable Machine Learning for Predicting Dental Treatment Complexity in Patients with Special Needs and Dentophobia. In: IEEE 2nd International Conference on AI and Emerging Technology For Sustainable Future, July 24-25, 2026, Catania, Italy. (In Press)
Preview |
Text
Final paper_juli.pdf - Accepted Version Download (291kB) | Preview |
Abstract
Dental treatment in patients with special needs and severe dentophobia represents a significant clinical and organizational challenge, often requiring individualized planning, including treatment under general anesthesia. This study proposes an explainable machine-learning framework for predicting the complexity of dental treatment using clinical, behavioral, demographic, medical, and socioeconomic data from 303 patients treated at a specialized university dental institution. Treatment complexity was modeled using a predefined binary TARGET variable, with higher-complexity cases representing patients requiring a greater procedural burden according to the clinical database. Logistic Regression, Random Forest, and XGBoost were evaluated using stratified 5-fold cross-validation with ROC-AUC, PR-AUC, and Brier score used to assess discrimination, high-risk case identification, and calibration. Permutation Feature Importance and SHAP analysis were applied to support model interpretability. Logistic Regression was selected as the final model, achieving ROC-AUC = 0.8828 and PR-AUC = 0.9014 with high interpretability. The most influential predictors were primary tooth extractions, permanent tooth extractions, restorative interventions, age, and neurodevelopmental disorders. Subgroup analysis indicated relatively stable performance across sex and financial-status groups, while regional estimates were less stable due to sparse categories. The results suggest that explainable machine learning can support transparent complexity stratification and resource planning in special-care dentistry. However, prospective validation using strictly pre-treatment variables is required before clinical deployment.
| Item Type: | Conference or Workshop Item (Paper) |
|---|---|
| Subjects: | Scientific Fields (Frascati) > Natural sciences > Computer and information sciences Scientific Fields (Frascati) > Engineering and Technology > Electrical engineering, electronic engineering,information engineering |
| Divisions: | Faculty of Information and Communication Technologies |
| Depositing User: | Prof. d-r. Blagoj Ristevski |
| Date Deposited: | 29 Oct 2026 19:52 |
| Last Modified: | 29 Oct 2026 19:52 |
| URI: | https://eprints.uklo.edu.mk/id/eprint/11940 |
Actions (login required)
 |
View Item |