Review finds fragmented reporting on bias in commercial medical AI validation
The recent flood of artificial intelligence (AI) models and content has undoubtedly introduced an unwelcome flaw from the human-made content it scrapes: bias. AI has been shown to form, copy and amplify harmful stereotypes of already marginalized groups. This phenomenon has far-reaching effects as AI becomes a normal part of everyday life.
A recent study has uncovered significant gaps in the reporting of demographic subgroup data for commercially available medical AI models, particularly in radiology. The review, led by Dr. Shannon L. Walston from Osaka Metropolitan University's Graduate School of Medicine, examined 545 studies on 252 AI products, but only 77 of these studies provided demographic data and subgroup performance results.
The researchers found that even within these studies, 67% of the datasets for tuberculosis detection did not have sufficient statistical power to analyze performance across sex subgroups. This lack of thorough reporting makes it difficult to assess the potential bias and ensure equitable clinical performance of these AI tools, which could directly impact patient care.
The findings, published in European Radiology, call for improved, transparent reporting standards across all stakeholders, including researchers and regulatory agencies, to build trust in medical AI products among physicians and patients.
Written by urgent.news from Medical Xpress's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
