@article{article_1520747, title={Machine Learning Based Software to Predict Type of Gingival Recession Surgery}, journal={European Annals of Dental Sciences}, volume={52}, pages={1–9}, year={2025}, DOI={10.52037/eads.2025.0001}, author={Karagöz, Burak and Bakırarar, Batuhan and Önder, Canan and Ünsal, Elif and Tatlıcıoğlu, Mehmet}, keywords={Gingival recession, Mucogingival surgery, Decision trees, Machine learning, Predictive decision model}, abstract={ABSTRACT
Purpose: The goal of this study is to identify the most important variables affecting gingival recession and to develop a machine learning-based software using these variables. Materials & Methods: 132 mandibular/maxillary right/left teeth #1, #2, #3, #4, and #5 were included in the study. Recession depth, recession width, width of keratinized gingiva, plaque index, buccogingival tissue thickness, frenulum position and mobility were recorded before and 6-month after surgery. One of the following methods has been selected: Gingival unit graft, coronally advanced flap, coronally advanced flap+connective tissue graft, and coronally advanced flap+platelet-rich fibrin. Software was developed to predict the type of gingival recession surgery.
Results: While the gingival unit graft group had the highest pre- and post-recession depth values, the coronally advanced flap group had the highest pre-recession width, pre-width of the keratinized gingiva and post-width of the keratinized gingiva values. A significant difference was found between surgical type categories for all variables except gender, post-buccogingival tissue thickness and post-frenulum position (p<0.05). Random Forest was found to be the best performing method both for surgery categories and overall based on accuracy and F-measure. Accuracy value was found 90.0% for gingival unit graft, 62.5% for coronally advanced flap, 71.4% for coronally advanced flap+connective tissue graft, and 97.8% for coronally advanced flap+platelet-rich fibrin. Conclusions: The machine learning software could evaluate the data accumulated in the database using the decision trees method and predict the prognosis of surgical techniques to treat gingival recession. The software developed will help physicians determine the optimal treatment approach.
Key Words: Gingival recession, Mucogingival surgery, Decision trees, Machine learning, Predictive decision model.}, number={1}, publisher={Ankara University}, organization={none}