Review finds fragmented demographic reporting in commercial AI validation

A comprehensive scoping review published in European Radiology reveals significant gaps in how commercial artificial intelligence tools for medical imaging are validated and reported, particularly regarding demographic bias. Led by Dr. Shannon L. Walston at Osaka Metropolitan University, the research analyzed 545 published studies assessing 252 commercially available radiology AI products. Despite growing concerns that algorithmic bias could exacerbate health disparities and compromise patient outcomes, the review found that demographic subgroup reporting remains critically fragmented. Only 77 of the 545 studies provided detailed demographic data or subgroup performance analysis, encompassing just 52 distinct products. When the research team applied statistical confidence intervals to datasets validating AI models for tuberculosis detection, they discovered that 67 percent of the reported data were underpowered for meaningful sex subgroup analysis. This lack of statistical rigor undermines the ability to accurately assess whether AI systems perform equitably across different populations. The findings underscore a systemic disconnect between the clinical demand for transparent, bias-resilient AI validation and the current state of commercial product testing. The authors emphasize that fragmented reporting patterns prevent regulators, clinicians, and developers from reliably estimating subgroup-specific bias. Without standardized, comprehensive demographic disclosures, it remains difficult to ensure that medical AI tools deliver safe and equitable outcomes across diverse patient groups. Dr. Walston and her colleagues note that resolving this issue requires coordinated action across the entire healthcare technology ecosystem. Researchers must adopt rigorous validation frameworks, commercial developers must prioritize inclusive dataset construction, and regulatory agencies must enforce transparent reporting standards. As AI integration into diagnostic radiology accelerates, establishing robust, demographically aware validation protocols is essential to maintaining clinical trust and patient safety. The study calls for immediate industry-wide reforms to align commercial AI validation with ethical and epidemiological best practices, ensuring that algorithmic decision support enhances rather than compromises equitable healthcare delivery.