Stanford researchers use AI for groundbreaking analysis of sex differences in brain structure

The scientists developed an AI system that accurately identified gender based on brain scans in 90% of cases. "The main motivation for conducting this study was the understanding that sex influences brain development, its aging, and symptoms of psychiatric and neurological disorders," explains Prof. Vinod Menon, the author of the study published in "Proceedings of the National Academy of Sciences" (doi: 10.1073/pnas.2310012121).

The AI system identified distinct differences in several aspects of the brain, including the default network responsible for introspection and self-reflection, the striatum involved in movement control, and the limbic system involved in triggering emotions and instinctual behaviors.

The debate about the influence of sex on brain organization has been ongoing. While we know that the brain is susceptible to sex-linked hormones, various brain structures appear the same in men and women and interact similarly.

However, the artificial intelligence system explored in the study seems to have discovered something that had previously eluded scientists. The scientists trained the system with brain scans obtained through magnetic resonance, concurrently indicating the sex of the person whose brain was being scanned. Over time, the computer model started to independently identify subtle differences unseen by other, previously used methods.

In subsequent tests involving 1.5 thousand scans, the system achieved 90% accuracy. The participants' brain scans presented to the computer came from different parts of the USA and Europe, suggesting that the discovered trends are universal.

"This is strong evidence that sex is a major factor determining the organization of the human brain," claims Prof. Menon, commenting on the study results.

The scientists employed technology known as explainable AI. Until recently, AI systems operated somewhat like a black box - it was unclear what information they processed or what they focused on. This approach enabled identification of neural networks within the brain that differed the most based on sex.

The researchers pushed further and developed additional models, adjusted for specific sexes, to investigate if brain scan results could predict outcomes in intellectual ability tests.

The results suggest that the previously observed differences between the sexes impact various aspects of mental efficiency. "These models work incredibly well because we included sex differences. This shows that ignoring these differences in brain organization might lead to overlooking important factors underlying neuropsychiatric disorders," emphasizes Prof. Menon.

Scientists believe that this same approach could be applied in a broader context. For instance, we could explore the relationships between the organization of different brain structures and varying cognitive abilities or behaviors. The researchers plan to make their technology accessible to all interested specialists.

"The AI models we developed have a broad range of applications. They can be employed to investigate how differences in brain structure influence learning disorders or social functioning. This is much-needed knowledge that will permit better support for individuals dealing with such challenges," highlights Prof. Menon.