Implementation of machine learning models as a quantitative evaluation tool for preclinical studies in dental...

Area of Science:

• Dental Education
• Haptic Skill Assessment
• Machine Learning in Dentistry

Background:

• Developing objective and reliable methods for evaluating haptic skills is crucial in dental education.
• Current evaluation methods may lack consistency, impacting skill development.

Purpose of the Study:

• To investigate the validity and reliability of a machine learning (ML) approach for assessing dental students' haptic skills.
• To compare ML-based evaluations with traditional examiner assessments.

Main Methods:

• 150 dental students performed Class II amalgam (C2A) and composite resin restorations (C2CR).
• Direct Observation Practical Skills forms and expert grading were used.
• A Python program with the Structural Similarity algorithm analyzed standard photographs of restorations.
• Validity and reliability were assessed using Cronbach's Alpha and Kappa statistics.

Main Results:

• The Structural Similarity Index (SSIM) demonstrated high intra-examiner reliability (α=0.961 for C2A, α=0.856 for C2CR).
• Compatibility between SSIM and examiner grades was statistically insignificant (p>0.05).
• Observer concordance was almost perfect for C2A (κ=0.806) and acceptable for C2CR (κ=0.769).
• Significant differences were noted in grading specific surfaces (occlusal for C2A, palatal for C2CR).

Conclusions:

• Machine learning, specifically the SSIM algorithm, is a promising tool for evaluating haptic skills in dental restorative procedures.
• Further improvements and alignment are necessary for fully objective and reliable ML-based validation in all dental training scenarios.
• ML offers a potential avenue for consistent and objective feedback in dental education.