Variation in standards for reporting artificial intelligence research using burn image datasets raises ethical...

Area of Science:

• Artificial intelligence
• Computer vision
• Medical image analysis

Background:

• Accurate burn depth and area assessment is crucial for treatment and prognosis.
• Current assessment methods rely on subjective visual analysis.
• Computer vision (CV), a sub-field of artificial intelligence (AI), offers potential for objective burn assessment.

Purpose of the Study:

• To conduct a scoping review of CV studies in burn care.
• To identify current practices in AI research involving burn image datasets.
• To determine if further guidance for researchers is needed.

Main Methods:

• Scoping review of studies from PubMed, arXiv, and IEE Xplore.
• Quantitative and qualitative analysis of 75 selected papers.
• Identification of common themes and reporting standards.

Main Results:

• 75 out of 1562 identified papers met inclusion criteria.
• Significant variation in reporting standards across studies.
• Common ethical issues include unavailable data/code, missing consent/ethical review statements, and insufficient dataset information.

Conclusions:

• Omissions in study reporting raise ethical questions and highlight the need for standardized reporting.
• A preliminary checklist for reporting burn image datasets in AI research is proposed.
• Standardized reporting will support researchers in this interdisciplinary field.