A new neural model involving artificial intelligence (AI) has shown itself capable of identifying a person’s gender using brain scans with an astonishing success rate of more than 90 percent. This discovery has revealed gender-based differences in brain anatomy, because those differences are so prominent in brain scans, it also upends one of science’s great debates.
This new research improves our understanding of brain development and ageing and also holds the potential to address gender-specific vulnerabilities to psychiatric and neurological disorders.
This AI model can identify males versus females with an accuracy rate above 90 percent on different brain scans, revealing apparent differences in how men’s and women’s brains are normally organised.
Because of explainability, researchers found that the default mode network, striatum, and limbic network are heavily involved in the AI model’s ability to distinguish between scans from people of different genders and, therefore, are likely critical to understanding brain function underlying behaviors.
These findings reinforce the need to consider sex differences in brain organisation for developing targeted interventions for neuropsychiatric disorders, a promising avenue for personalised medicine.
A team of neuroscientists at Stanford Medicine recently showed that an AI model has achieved greater than 90 percent accuracy in determining whether an MRI scan of the brain looked like one of a man or a woman.
The results, published 19 February in the Proceedings of the National Academy of Sciences, mark the end to decades-long debates over the validity of gender-based disparities in the human brain – and, scientists argue, represent a necessary first step towards tackling neuropsychiatric disease in both sexes.
Dr. Vinod Menon, professor of psychiatry and behavioural sciences and director of the Stanford Cognitive and Systems Neuroscience Laboratory, said: “A key motivation for this study is that sex plays a crucial role in human brain development, in aging, and in the manifestation of psychiatric and neurological disorders,” he added: “Identifying consistent and replicable sex differences in the healthy adult brain is a critical step toward a deeper understanding of sex-specific vulnerabilities in psychiatric and neurological disorders.”
Despite much-preceding research failing to identify consistent brain variables for gender, the availability of data and advanced artificial intelligence allowed Menon’s team to run a more robust analysis.
They had the highest success in revealing the role of gender at organising the human brain ever reported, using a deep neural network model to predict the gender of the scanned brain from each frame of a dynamic MRI. Using this approach, researchers disentangled inter-site variance – such as the scanning equipment and techniques – from the brain structure to gender variance, finally offering clear evidence of the role of sex in structuring the human brain.
Furthermore, by leveraging explainable AI, the team explained the specific networks in the brain on which the model was relying to make that decision, setting the stage for further research that studies the implications of gendered differences in cognitive function and behaviour.
Such models are similarly universal and, if Menon is successful in determining which nodes in the default model are associated with different cognitive functions or behaviours, it could open the way to some fascinating ideas about what goes wrong in cases of learning disorders or variances in social functioning. The new model will be made available to anyone who wants to try it out.
Source: Erin Digitale, Stanford,neurosciencenews.com February 21, 2024
