BIAS: Transparent reporting of biomedical image analysis challenges

Area of Science:

• Biomedical image analysis
• Computational pathology
• Medical imaging informatics

Background:

• The number of annual biomedical image analysis challenges is rising globally.
• Current challenge reporting practices often lack sufficient detail for adequate interpretation and reproducibility of results.
• This gap impacts the reliability and comparability of algorithmic benchmarking in the field.

Purpose of the Study:

• To address the shortcomings in reporting standards for biomedical image analysis challenges.
• To introduce the Biomedical Image Analysis ChallengeS (BIAS) initiative and its developed recommendations.
• To enhance the transparency, interpretability, and reproducibility of challenge outcomes.

Main Methods:

• Development of the BIAS statement through a structured initiative.
• Creation of a comprehensive checklist for authors submitting challenge-related papers.
• Focus on improving reporting regardless of application, image modality, or task.

Main Results:

• Establishment of a set of recommendations for reporting biomedical image analysis challenges.
• Introduction of a practical checklist to guide authors and reviewers.
• The BIAS statement aims to standardize reporting across diverse challenges.

Conclusions:

• The BIAS statement and checklist are crucial for improving the quality of biomedical image analysis challenge reporting.
• Standardized reporting will facilitate better interpretation and reproducibility of challenge results.
• Adoption of these guidelines is encouraged to advance the field of biomedical image analysis research.