ASSESSMENT OF DENTURE BORDER EXTENSION UTILIZING ARTIFICIAL INTELLIGENCE: AN AUTOMATED DETECTION MECHANISM FOR ENHANCED PREDICTIVE PRECISION

Abstract

Introduction

Effective complete denture treatment relies heavily on the correct extension of the denture base, which ensures healthy soft tissue and optimal retention. Inaccurate extensions can compromise functionality and aesthetics, making long-term success challenging. While traditional border extension assessments are subjective and time-intensive, machine learning offers a more objective approach. Using AI, clinicians can quickly and precisely evaluate the denture base's extension quality, enhancing the efficiency of dental labs. Though initial machine learning results are promising, further research is needed, especially in refining algorithms for various digital imaging methods. This study aims to evaluate AI's role in distinguishing between well-fitted and ill-fitted complete denture base extension.

Background And Aim

In this research, we explored the use of two separate machine-learning techniques to automatically identify the extension of the denture base.

 Materials and Methods

In the study, the machine learning platform, Orange, employed an algorithm based on SqueezeNet embeddings. Logistic regression and Naïve Bayes techniques were utilized to forecast and identify these algorithms. These models underwent training and evaluation on 40 diverse images showcasing various border molding scenarios. Their precision was then gauged using a confusion matrix.

Results

Utilizing CNN embeddings, three machine learning models effectively identified and anticipated the denture base extension. Each of these models demonstrated notable accuracy in their predictions. Specifically, Naïve Bayes exhibited an AUC of 0.982, while Logistic Regression achieved a perfect AUC score of 1.000.

Conclusion

Machine learning models exhibit a strong capability to differentiate and categorize denture bases according to their extension quality, whether deemed satisfactory or unsatisfactory, with notable precision. To validate and enhance these observations, further research encompassing larger datasets and diverse imaging techniques is essential.

Keywords : Deep Learning, Artificial intelligence, Denture base, Border molding