The abstract is read at the publisher; the summary is JClub's.
A physics-informed neural network reconstructs cellular membrane geometries to curvature-order accuracy from 3D electron microscopy images, with physics-based regularization improving curvature recovery under image noise.
- Why it matters: Reliable curvature estimation is needed to quantify membrane ultrastructure and link protein organization to structure and function, but higher-order derivatives from discrete surfaces are difficult.
- What they did: The method uses an implicit neural representation trained with physics-based regularization, benchmarked on synthetic data and then applied to experimental membranes including the Golgi apparatus, mitochondria, and endoplasmic reticulum–plasma membrane contact sites.
- The result: It generalizes to complex structures and captures localized deformations at contact sites; curvature analysis of endocytic pits provides preliminary evidence for anisotropic curvature at the pit neck consistent with a lower-energy constriction pathway.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.