Deep learning models reveal replicable, generalizable, and behaviorally relevant sex differences in human functional...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Genetics

Background:

• Sex significantly influences brain development, aging, and disease manifestation.
• Previous studies on sex differences in brain organization yielded inconsistent and unreproducible findings.

Purpose of the Study:

• To develop a novel AI model to identify reliable sex differences in human functional brain organization.
• To investigate the behavioral relevance and generalizability of these sex differences.

Main Methods:

• A spatiotemporal deep neural network (stDNN) was employed to analyze functional brain dynamics.
• Explainable AI (XAI) techniques were used to identify key brain features distinguishing sexes.
• The model was validated across multiple sessions and three independent cohorts (N ≈ 1,500).

Main Results:

• The stDNN model achieved high accuracy (>90%) in differentiating male and female brains, demonstrating strong replicability and generalizability.
• XAI identified significant sex differences in the default mode network, striatum, and limbic network.
• XAI-derived brain features predicted sex-specific cognitive profiles with high accuracy.

Conclusions:

• Functional brain organization exhibits robust, generalizable, and behaviorally relevant sex differences.
• These findings challenge the concept of a male-female brain continuum and highlight sex as a critical biological factor.
• The study provides AI-driven tools for personalized, sex-specific biomarker development in neurological and psychiatric disorders.