Predicting brain morphogenesis via physics-transfer learning.

Physics-transfer learning enables accurate prediction of brain morphogenesis with limited labeled data, achieving strong generalization across complex biological systems.

• Why it matters: Understanding brain development is hindered by the intricate folding patterns and scarce labeled data, which challenge traditional modeling approaches and limit insights into neurodevelopmental processes.
• What they did: The authors developed a physics-grounded neural network framework that embeds nonlinear elasticity laws into models and transfers this knowledge across different physical domains, deriving a generalization bound.
• The result: This approach accurately characterizes brain features and predicts morphogenesis, providing reduced-dimensional representations that capture essential physics, supporting digital-twin technologies for brain health assessment.