Brain volumetric variability and artificial intelligence diagnosis: Importance of race/ethnicity-specific reference standards and social determinant adjustment. A scoping review

Neuroimaging techniques such as magnetic resonance imaging (MRI) are routinely used in diagnostic radiology to evaluate brain changes associated with neurological and psychiatric conditions. Evidence suggests that imaging biomarkers predict clinical outcomes with varying accuracy across ethnic groups. Underrepresentation of ethnic diversity in neuroimaging research may result in generalization bias, whereby findings derived from one population are inaccurately applied to others. A scoping review methodology was employed to systematically identify and analyze relevant literature. Searches were conducted across EBSCOhost (including CINAHL and Medline Complete), Elsevier (Scopus), Clarivate (Web of Science), and PubMed. Eligible studies examined ethnicity-related differences in subcortical brain volumes and cortical thickness in healthy adults using neuroimaging. The search yielded 1,013 records, which were screened according to predefined inclusion and exclusion criteria. Fourteen studies met the eligibility criteria and were included in the final analysis. The reviewed studies demonstrated significant variations in cortical thickness and subcortical volumes across diverse populations and socioeconomic groups, underscoring the need for population-sensitive reference standards in neuroimaging to minimize generalization bias. These findings highlight the importance of incorporating ethnic variability into neuroimaging research and developing population-sensitive frameworks for MRI-based diagnostic applications. Additionally, the review identifies key areas for improvement, including the integration of ethnic and socioeconomic diversity in artificial intelligence (AI)-driven neuroimaging models to enhance diagnostic precision and equity.