Machine learning prediction of marginal gingival adaptation of direct composite veneering with or without mock-up in Upper anterior

This research uses machine learning algorithms to automate the detection and adaptation of marginal gingival in direct composite veneering. This approach improves treatment predictability and refines protocols, thereby contributing to the evolution of aesthetic dentistry practices. The study aims to enhance diagnostic accuracy. The application of distinct machine-learning algorithms for automated detection and marginal gingival adaptation of direct composite veneering with or without mock-up in the upper anterior was investigated in this study. A machine learning tool, orange, was used with an algorithm using squeeze net embedding models. Logistic Regression and naïve bayes were used to predict and classify based on deep learning embeddings, which were trained and tested on 341 images of direct composite veneering with and without mock-up in upper anteriors and evaluated for accuracy. Logistic Regression outperforms Naive Bayes in predicting marginal gingival adaptation in direct composite veneering. It has a higher accuracy rate and precision in identifying positive cases, minimizing false positives, and identifying cases where good marginal gingival adaptation occurs, thereby enhancing its effectiveness in predicting dental procedures. Based on the marginal gingival adaptation of direct composite veneering with or without mock-up in the Upper anterior, machine learning algorithms can identify the marginal adaptation of composite veneering with and without mock-up with high accuracy. More studies with bigger sample numbers and additional imaging modalities are needed to corroborate these findings.