Unveiling Brain Sex Differences with AI

In a groundbreaking study, Stanford Medicine investigators have developed an artificial intelligence model that boasts over 90% accuracy in determining whether brain activity scans belong to a woman or a man. This significant achievement, published on February 20 in the Proceedings of the National Academy of Sciences, offers a resolution to the longstanding debate on the existence of reliable sex differences in the human brain. It also underscores the importance of these differences in addressing neuropsychiatric conditions that distinctly affect women and men.

“Human brain illustrated with millions of small nerves” by Ars Electronica is licensed under CC BY-NC-ND 2.0

Vinod Menon, PhD, the senior author of the study, emphasizes the importance of sex in brain development, aging, and the manifestation of psychiatric and neurological disorders. Menon, along with lead authors Srikanth Ryali, PhD, and Yuan Zhang, PhD, has taken a step forward in identifying consistent and replicable sex differences in the healthy adult brain. This is seen as a critical move towards understanding sex-specific vulnerabilities in psychiatric and neurological disorders.

The AI model’s success is attributed to its analysis of dynamic MRI scans, which capture the complex interactions among different brain regions. The ‘hotspots’ that most aided the model in distinguishing between male and female brains include the default mode network, which processes self-referential information, and the striatum and limbic network, which are involved in learning and response to rewards.

The study does not take a stance on whether these sex-related differences are innate from early life or are influenced by hormonal differences or societal circumstances encountered by men and women. However, the existence of these differences is clear, and the AI model’s ability to detect them with such high accuracy is a testament to the robustness of sex as a determinant of human brain organization.

The researchers utilized ‘explainable AI’ to identify the brain networks that were pivotal to the model’s decisions. This tool allows for a deeper understanding of how the AI model makes its judgments, which in this case, pointed to the default mode network, striatum, and limbic network.

Furthermore, the team explored the possibility of predicting cognitive task performance based on functional brain features that differ between sexes. They developed sex-specific models that effectively predicted cognitive performance, indicating that functional brain characteristics varying between sexes have significant behavioral implications.

The research has wide implications, advancing brain understanding and opening new avenues for personalized medicine by recognizing sex differences. Supported by the NIH, the study will be published in PNAS. Menon and team will share their AI model for broader research use.

This study by Stanford Medicine investigators marks a significant stride in neuroscience, offering compelling evidence of sex differences in brain organization and opening doors to personalized approaches in medicine.