AI finds women and men are different when it comes to brain organization and function

Researchers at Stanford Medicine have reported on the development of a new artificial intelligence model that was found to be more than 90% successful in testing whether MRI scans of human brain activity come from men or women. The researchers suggest that the findings will help resolve a long-standing debate about whether reliable sex differences exist in the human brain, and that understanding these differences will help scientists It shows that it may help us better understand neuropsychiatric conditions that affect women and men differently.

“A key motivation for this study is that sex plays an important role in human brain development, aging, and the manifestation of psychiatric and neurological diseases,” said Professor of Psychiatry and Behavioral Sciences and director of Stanford University. said Dr. Vinod Menon. Cognitive and Systems Neuroscience Laboratory “Identifying consistent and reproducible sex differences in the healthy adult brain is an important step toward a deeper understanding of sex-specific vulnerabilities in psychiatric and neurological diseases .”

Menon is the senior author of the study published by the team. PNAS” with the title,Deep learning models reveal reproducible, generalizable and behaviorally relevant sex differences in human functional brain organization“Our findings highlight the important role of gender as a biological determinant in human brain organization, and the role of individualized sex in psychiatric and neurological diseases,” the researchers wrote in their report. “This has important implications for the development of gender-specific biomarkers and provides an innovative AI-based computational tool.” For future research. ” The lead authors are Dr. Srikanth Ryali and his academic staff researcher Dr. Yuan Zhang.

The authors noted that sex plays an important role in early brain development, puberty, and aging. Furthermore, they argue that “gender is an important biological factor that influences human behavior, influencing brain function and the expression of psychiatric and neurological diseases. Therefore, our knowledge of sex differences in the human brain is essential for understanding both normative behavior and psychopathology.'' Indeed, the extent to which a person's gender influences how the brain is organized and operates has long been a matter of debate among scientists. Scientists suggested that there is.

We know that the sex chromosomes we are born with help determine the cocktail of hormones our brains are exposed to, especially during early development, adolescence, and aging, but researchers has long struggled to link gender to specific differences in the human brain. Brain structures tend to look similar in men and women, and previous studies of how brain regions work together have rarely revealed consistent indicators of the brain's gender. Not yet. “…previous studies on how brain organization differs between men and women have been inconclusive,” the researchers noted. “…our understanding of sex differences in the functional organization of the human brain and their behavioral implications has been hampered by inconsistent findings and lack of replication.”

In the newly reported study, Menon and colleagues exploited recent advances in artificial intelligence and access to multiple large datasets to pursue more powerful analysis than previously employed. First, we created an end-to-end spatiotemporal deep neural network (stDNN) model and trained it to classify data from Human Connectome Project (HCP) resting-state functional MRI (rsfMRI) brain images. When researchers showed models brain scans and told them they were looking at a male or female brain, the models began to “notice” subtle patterns that helped them tell the difference. “Our stDNN model revealed reliable sex differences with cross-validation classification accuracy of over 90%, outperforming previous studies,” the scientists said.

The team evaluated the reproducibility of the predictive model on additional datasets without further training. When tested on about 1,500 brain scans, the stDNN model was almost always able to tell whether a scan was from a woman or a man. It also used a deep neural network to analyze dynamic rsfMRI scans, which showed superior performance compared to models used in previous studies. This approach captures the complex interactions between different brain regions. “Importantly, our model outperformed previous work both in testing and on independent datasets,” the team said.

The success of this model suggests that detectable sex differences exist in the brain but have not been reliably detected until now. The fact that this model performed so well on a variety of datasets, including brain scans from multiple centers in the United States and Europe, while controlling for the many confounding factors that can influence this type of research, This makes this result particularly convincing. “This is very strong evidence that sex is a strong determinant of human brain organization,” Menon said.

Until recently, models like the one adopted by Menon's team have helped researchers classify brains into different groups, but they cannot provide information about how the classification is done. was. But researchers today have access to a tool called Explainable AI (XAI) that can sift through vast amounts of data and explain how a model's decisions are made.

Menon and his team used explainable AI to identify the brain networks that were most important in helping the model determine whether a brain scan was from a man or a woman. They found that the “hot spots” that most help the model distinguish between male and female brains include the default mode network (DMN), a brain system that helps process self-referential information; We found that the striatal and limbic networks are involved. How we respond to learning and rewards.

Remarkably, the DMN, striatal, and limbic networks contribute to the prevalence of conditions such as autism, attention deficit disorder, depression, addiction, schizophrenia, and Parkinson's disease in women or men. It is also a site of dysfunction in mental illness that is disproportionate,” they note. It is associated with gender-specific sequelae and outcomes. ” The research team suggested that their findings “…may therefore provide a template for investigating gender-specific vulnerability to individual psychiatric and neurological disorders.”

The researchers also wondered if they could create another model that could predict how well participants would perform certain cognitive tasks, based on the different functional characteristics of the brain between women and men. To do this, they developed a gender-specific cognitive ability model. One model effectively predicted the cognitive ability of men but not women, and the other effectively predicted the cognitive ability of women but not men. The results of this study showed that brain functional characteristics that differ between the sexes have a significant impact on behavior. “Importantly, the brain features identified by XAI that reliably differentiate functional brain organization between the sexes also predicted unique cognitive profiles for women and men,” the authors commented.

Summarizing their findings, they concluded: “Our approach using spatiotemporal DNN and XAI techniques identifies reproducible, generalizable, and interpretable sex differences in human functional brain organization across multiple datasets and independent cohorts; We reveal that brain functional characteristics differ between set and independent cohorts; gender is behaviorally important.”

Menon added, “These models worked so well because they were able to separate brain patterns between men and women. We can see how important factors can be missed.”

While the research team applied deep neural network models to questions about gender differences, Menon said the model could be used to answer questions about how nearly every aspect of brain connectivity relates to all kinds of cognitive abilities and behaviors. states that it can be applied to answer. The scientists plan to make their model publicly available for any researcher to use. “Our AI model has very broad applicability,” Menon says. “Researchers can use our model to look for brain differences associated with learning disabilities or differences in social functioning, for example, to help individuals adapt to and overcome these challenges. This is an aspect that we would like to understand better.”

The researchers noted that the study did not address whether sex differences arise early in life or are caused by hormonal differences or differences in the social situations men and women are more likely to encounter. Nevertheless, they write: “The discovery of robust brain functional features underlying sex differences has the potential to inform quantitatively accurate models for investigating sex differences in psychiatric and neurological disorders. , paving the way for more targeted and personalized approaches in both cognitive neuroscience research and clinical applications.”