How should studies using AI be reported? lessons from a systematic review in cardiac MRI

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence
• Medical Informatics

Background:

• Artificial intelligence (AI) tools are increasingly used for cardiac MRI (CMR) segmentation.
• High-quality reporting is essential for the advancement and clinical adoption of these AI tools.
• Previous reviews have highlighted variable reporting standards in AI medical imaging research.

Purpose of the Study:

• To systematically evaluate the reporting quality of studies using AI for cardiac MRI segmentation.
• To identify common deficiencies in study reporting.
• To provide recommendations for improving future research reporting in this domain.

Main Methods:

• A systematic review was conducted assessing 209 studies.
• Studies were evaluated for compliance with the Checklist for AI in Medical Imaging (CLAIM).
• Reporting of model descriptions, study design, datasets, and performance was specifically assessed.

Main Results:

• Compliance with CLAIM was high for model descriptions (100%) but lower for study design (71%), datasets (63%), and performance (60%).
• Significant and frequent information gaps were identified across publications.
• Overall reporting quality was variable and sometimes poor.

Conclusions:

• The quality of reporting in AI for cardiac MRI segmentation studies needs substantial improvement.
• Adherence to reporting guidelines like CLAIM is crucial for transparency and reproducibility.
• Enhanced reporting will facilitate the critical appraisal and clinical translation of AI tools in cardiology.