Frozen neuroimaging models struggle with african brain mri data
Evaluating the Generalization of Neuroimaging Foundation Models on African Brain MRI
Computer Vision and Pattern Recognition
Summary
Pretrained brain MRI models built using mostly western data do not work well when tested on brain scans from Nigerian patients. The authors tested several popular models to see if they could accurately classify different diseases but found most just guessed the most common category. A specially trained model on the Nigerian data did better but still has limited accuracy. This suggests these existing models need changes before they can be fairly used in different populations around the world.
What this means in practice
- •For clinical ai developers: Improve brain MRI diagnostic tools for underrepresented populations by integrating adaptation methods suited for small non-western datasets.
- •For medical imaging device companies: Develop more inclusive brain MRI software that maintains accuracy across diverse global patient populations using validation techniques from this study.$Commercial implications: Enables creation of brain MRI diagnostic software for African markets addressing unmet generalization needs.
Authors
Oluwatobi Iyanuoluwa Akinmuleya, Olatokun Shamsudeen Akano, Samuel Danquah Ankapong, Olamide Lawal, Toufiq Musah
Abstract
Neuroimaging foundation models pretrained on large, predominantly western cohorts are increasingly proposed as general-purpose backbones for brain MRI analysis. Yet, their ability to generalize to underrepresented clinical populations remains largely untested. We evaluate four recent foundation models (BrainIAC, Neuro-JEPA, NeuroVFM, and Primus) on a three-way diagnostic classification task (Control, Dementia, Parkinson's disease) using a cohort of 88 subjects from a Nigerian clinical brain MRI dataset, across four modality configurations (T1w, T2w, T1w+T2w, FLAIR), and compare against an end-to-end trained ViT3D baseline. The frozen backbones collapse to majority-class predictions, while Neuro-JEPA on FLAIR shows modest but still limited discrimination. In contrast, the end-to-end trained ViT3D achieves higher accuracy and MCC on every task (up to 53.4% accuracy, MCC=0.27) and is the only model with non-trivial recall. Our findings suggest that these frozen neuroimaging foundation models are insufficient for fine-grained diagnostic classification in small, non-western clinical cohorts, motivating parameter-efficient adaptation and broader multi-site external validation for equitable deployment in global health settings.