A literary review of machine learning sex classification methods

Being able to determine a person’s biological sex from brain scans offers valuable insight into how male and female brains differ in structure and function. These differences are linked to variations in how certain neurological and psychiatric disorders, such as Alzheimer’s disease, autism spectrum disorders, and schizophrenia, develop and progress. Traditional methods for sex classification have often relied on comparing overall brain volumes, which can miss subtle and complex patterns in brain connectivity and shape. Various technical advances in neuroimaging now make it possible to examine the brain from multiple perspectives, capturing both fine-grained structural details and dynamic functional activity. The scale and complexity of these data require more powerful analytical tools, and recent work has turned to machine learning (ML) and artificial intelligence (AI) models. For such models to be useful, they must not only achieve robust and generalizable performance but also provide meaningful insights into sex-related brain differences and their behavioral implications. In this review, we critically analyze recent AI-based studies according to criteria including model performance and generalizability, uni- versus multi- modal approaches, identification of biomarkers, and brain–behavior associations. We further discuss the relative strengths and limitations of different methods within these frameworks.