AI can identify a child’s sex based on their brain activity

Artificial intelligence can differentiate between the brain patterns of boys and girls aged 9 to 10 years old according to their sex, and possibly their gender – but not everyone is convinced by the accuracy of the results.

The prevalence of conditions such as pain, headache and heart disease differs between the sexes, but we know little about the neurological variations here or between genders, particularly among children.

To learn more, Elvisha Dhamala at the Feinstein Institutes for Medical Research in New York and her colleagues analysed thousands of sets of magnetic resonance imaging (MRI) data from more than 4700 children, with a roughly even split between the sexes. The children were all aged 9 to 10 and are participating in the Adolescent Brain Cognitive Development project.

Sex was defined according to someone’s “anatomy, physiology, genetics and/or hormones at birth”. Gender was judged according to “features of an individual’s attitude, feelings and behaviours”.

Parents weren’t asked outright what they thought their children’s genders were. Instead, this was assessed by asking them a series of questions, such as how often their children imitate male or female TV and film characters, whether they state that they wish to be a girl or a boy, and if they say they dislike their genitals. All these questions were weighted equally and combined into a score.

A separate score was created from questions asked to the children themselves, such as whether they felt like a boy or a girl.

The researchers haven’t disclosed the different genders that the children may have identified with or how many of the children had a gender that differed from their sex. “Gender was considered on a continuum, not as a binary,” says Dhamala. “We did not reduce our analyses to categorical genders so we are unable to comment on how many children had a gender different to their sex.”

They first looked at associations between brain networks and sex, then between these networks and gender within each assigned sex. The team found that different sexes and genders are associated with distinct patterns of functional connectivity, a measure of how distant brain regions communicate.

Sex was associated with connectivity between the visual cortex, areas that control movement, and the limbic system, a group of deep brain structures involved in regulating emotions, behaviour, motivation and memory. These networks “were important to distinguish participants based on their sex”, says Dhamala.

The network associated with gender was more widely distributed throughout the cerebral cortex – the outer layer of the brain, which is also linked to memory and movement, as well as sensation and problem-solving. This was true when using the gender score created from the parents’ answers to their questions and also when using the separate score from questioning the children themselves.

“In assigned females, gender mapped onto networks involved in attention, emotional processing, motor control and higher-order thinking,” says Dhamala. “In assigned males, the same relationships were present, but there were also additional networks involved in higher-order thinking and visual processing. There was some overlap between the brain networks associated with sex and gender, but they were largely distinct from one another.”

After the researchers trained an AI model on some of this MRI data, it could identify a child’s sex based on the brain connectivity patterns within other sets of the data. It could also predict gender, but far less accurately than for sex and only according to the genders reported by the parents, not the children themselves.

Better understanding how brain activity patterns differ according to sex and gender could help scientists learn more about conditions that vary in prevalence between boys and girls, such as ADHD, says Dhamala.

The findings could also have implications for how human brain research is conducted, she says. “This tells us that we need to start considering sex and gender separately in biomedical research, and this holds true for how we collect data, analyse it, and also how we interpret and communicate our results,” says Dhamala.

But Ragini Verma at the University of Pennsylvania says the study tells us little about the neurological basis of gender. She says that the team probably only found signals of distinct brain activity patterns among different genders because of the study’s large sample size, but that “the variability in gender prediction is based on low accuracy”.