Researchers have succeeded in training self-learning artificial intelligence (AI) software to recognize whether brains shown in fMRI scans are male or female. This indicates that there are characteristic gender differences in the connections between regions of the brain. The scientists, who hail from Jülich, Düsseldorf, and Singapore, have published their findings in the journal Cerebral Cortex.
Countless scientific studies have attempted to determine whether men’s and women’s brains function differently. The results yielded by these studies often differ. In many cases, they are also a source of contention, as they are based on data obtained from small control groups and thus may be distorted by factors that are not immediately obvious. “In contrast, our methods – which utilize artificial intelligence – produce very reliable results,” says PD Dr. Susanne Weis, lead author of a new study that has just been published. Weis works at the Institute of Neuroscience and Medicine at Forschungszentrum Jülich as well as at the Institute of Systems Neuroscience at Heinrich Heine University Düsseldorf.
According to the study, the male and female brain show particular differences in functional connectivity between certain areas of the cingulate gyrus, the precuneus, and the medial frontal cortex. These networks play an important role in language function, emotional processing, and social perception.
For their study, the team of brain researchers led by Dr. Weis and Prof. Simon Eickhoff initially used brain scans from 434 test subjects from the Human Connectome Project. The brain scans – or, to be more precise, functional magnetic resonance imaging (fMRI) scans – reveal which areas of the brain are active and interacting at the moment the image is taken. While the scans were being taken, the test subjects let their minds wander. The researchers trained a self-learning software to categorize the test subjects’ gender on the basis of neural interaction patterns. Throughout the training process, they constantly gave the software feedback on the accuracy of its results, which enabled the AI to progressively improve its mathematical model. The scientists then used the AI software to predict the gender of a further 310 test subjects from the Human Connectome Project and 941 test subjects from the 1000BRAINS study, all on the basis of fMRI scans. It did so with an accuracy of approximately 70 %.
“The results of the study show that areas in women’s brains are networked and connected differently to men’s brains,” says Dr. Weis. However, she is keen to stress: “On no account can the results be interpreted as evidence for assessments such as ‘women are better at dealing with emotions.’” For now, the reasons for these differences in the brain remain an open question: possibilities include both biological and acquired causes, for instance as a result of children’s upbringing.
In their study, the researchers focused on a binary view of gender as an initial approach. In principle, however, their methodology could be used to investigate the extent to which intersexuality or transsexuality is reflected in the networks in the brain. “But to do so, we would need a much larger quantity of brain scans and data from intersex or transsexual people. The kind of comprehensive study this would require has not yet been undertaken,” says Weis.