Core outcomes measures in dental computer vision studies (DentalCOMS)

Area of Science:

• Dental research
• Artificial Intelligence
• Computer Vision

Background:

• Dental computer vision studies face challenges in reporting and comparability due to varied outcome measures.
• Bias and difficulty in synthesizing research hinder progress in the field.

Purpose of the Study:

• To develop a consensus-based Core Outcome Measures Set (COMS) for dental computer vision studies.
• To improve the quality, comparability, and reduce bias in reporting.

Main Methods:

• Reviewed existing diagnostic accuracy study guidance and conducted expert interviews.
• Mapped outcome measures to computer vision tasks and clinical fields.
• Utilized a two-stage e-Delphi consensus process with 26 international participants.

Main Results:

• Established agreed reporting levels for various computer vision tasks.
• Determined human expert assessment and diagnostic accuracy are crucial for meaningful evaluation.
• Identified eight core outcome measures: confusion matrix, accuracy, sensitivity, specificity, precision, F-1 score, AUROC, and AUPRC.

Conclusions:

• Dental researchers should adopt the defined COMS for computer vision studies.
• Reviewers and editors should consider the COMS for study assessment.
• Adherence to the COMS will improve study synthesis and reduce selective reporting.