Segmentation of process-related contaminations on two-piece abutments using pixel-based machine learning: a new...

Area of Science:

• Biomaterials Science
• Dental Materials
• Computational Biology

Background:

• Establishing a reliable method for quantifying contamination on CAD/CAM-manufactured two-piece abutments is crucial for dental applications.
• Current methods for assessing surface cleanliness lack standardization, necessitating novel approaches.

Purpose of the Study:

• To investigate a pixel-based machine learning (ML) method for detecting and quantifying contamination on customized two-piece zirconia abutments.
• To embed the ML method within a semiautomated pipeline for contamination analysis.

Main Methods:

• Forty-nine CAD/CAM zirconia abutments were analyzed using scanning electron microscopy (SEM) for contamination.
• Pixel-based ML and simple thresholding (SW) were employed for contamination detection and quantification.
• Statistical comparison using Wilcoxon signed-rank test and Bland-Altmann plots assessed method agreement.

Main Results:

• No statistically significant difference was found between ML and SW methods in measuring contamination area percentages (P = 0.22).
• The Bland-Altmann plot indicated good agreement, with a mean difference of -0.006% for ML compared to SW.
• ML showed potential for detecting contamination fractions above 0.03%.

Conclusions:

• Both segmentation methods, ML and SW, demonstrated comparable efficacy in evaluating surface cleanliness of zirconia abutments.
• Pixel-based ML presents a promising tool for detecting external contaminations on zirconia abutments.
• Further research is needed to evaluate the clinical performance of the ML-based assessment tool.