CLAIRE: a unified framework for reporting and assessing artificial intelligence in diagnostic imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Clinical Reporting

Background:

• Inconsistent reporting hinders reproducibility of artificial intelligence (AI) models in diagnostic imaging.
• Current guidelines lack specificity for AI diagnostics, particularly clinical usability and technical transparency.

Purpose of the Study:

• To develop and validate the Completeness, Learnability, Applicability, Interpretability, Reproducibility, and Evaluation (CLAIRE) framework.
• To create a practical reporting aid for AI in medical and dental imaging.

Main Methods:

• Retrospective validation of the CLAIRE framework on 10 imaging studies.
• Internal validation assessing inter-rater and intra-rater reliability.
• Development of a 15-item checklist, scoring system, and editorial guide.

Main Results:

• High inter-rater agreement improved from Cohen's κ 0.286 to 0.987 (p < 0.01) after calibration.
• Mean intra-rater reliability reached 0.997 after a six-month washout period.
• The framework yielded a structured checklist, scoring system, and guide for AI reporting.

Conclusions:

• The CLAIRE framework offers a unified structure to enhance reporting consistency for AI imaging models.
• It supports systematic appraisal, improving reproducibility and clinical translation.
• CLAIRE aims to increase clinician accessibility through plain-language summaries and applicability assessments.