Geometric Shape Descriptors for High Precision Medical Phenotyping

This direction is merged into another direction: surface-based morphometry. Its last version stays readable and citable here.

Version 1 of 1113 papers129 conceptsfrozen 2026-10-04

In clinical medicine, we often describe organs using "scalar" metrics: the volume of a hippocampus, the thickness of a cortex, or the mass of a heart. While these numbers are easy to calculate, they are "blunt instruments." They collapse complex, three-dimensional biological structures into a single value, discarding the vast majority of the spatial information.

A growing body of research suggests that the geometry of an organ—its curvature, folding patterns, and topological "skeleton"—contains far more diagnostic and predictive power than its size alone. By treating biological structures as manifolds rather than just containers of volume, researchers are uncovering "preclinical" signatures of disease that traditional imaging misses.

The Failure of Volumetric Averaging

The fundamental problem with volume is that it is an aggregate measure. If a brain region undergoes subtle remodeling—where one part expands while another contracts—the total volume might remain unchanged. This "averaging out" masks the early stages of neurodegeneration or developmental shifts.

As Martin Reuter and colleagues demonstrated with BrainPrint, localized morphological warping provides a much earlier signal of Alzheimer’s than total volume loss. Similarly, Jianyang Xie found that the roundness of the foveal avascular zone in the eye is a more sensitive marker for cognitive impairment than its total area Deep segmentation of OCTA for evaluation and association of changes of retinal microvasculature with Alzheimer’s disease and mild cognitive impairment. In these cases, the shape changes long before the size does.

Curvature as a Proxy for Biological Stress

To capture these subtle shifts, researchers borrow tools from differential geometry, specifically the concept of curvature. Curvature measures how a surface deviates from being flat. In a biological context, changes in curvature often signal underlying physiological stress or plastic reconfiguration.

Folding and Topology: The Brain’s Blueprint

The "wrinkliness" of the brain, or gyrification, is not random; it follows a blueprint established in the womb. Because these folding patterns are stable throughout life, they serve as a "fossil record" of early development.

Colleen Rollins found that specific deviations in the depth and fragmentation of cortical folds (sulci) can predict whether a patient with schizophrenia will experience hallucinations Evidence in cortical folding patterns for prenatal predispositions to hallucinations in schizophrenia. This suggests that the geometric configuration of the brain's surface is a marker for functional network impairments that volume cannot capture.

Furthermore, these geometric traits appear to be more "biological" than scalar ones. Tian Ge and Anderson M. Winkler demonstrated that multidimensional shape descriptors exhibit higher heritability than traditional volume metrics, meaning they are more closely tied to our genetic code Multidimensional heritability analysis of neuroanatomical shape.

Anchoring Shape with Skeletal Representations

One challenge in shape analysis is that the "outer surface" of an organ can be noisy or difficult to segment. To solve this, researchers like Yanwu Yang have developed frameworks like HippMetric, which uses an internal "skeletal" scaffold to map the hippocampus HippMetric: A skeletal-representation-based framework for cross-sectional and longitudinal hippocampal substructural morphometry.

By anchoring measurements to this stable interior scaffold rather than the volatile surface, the model can track "skeletal thickness" with high precision. This geometric approach significantly outperforms global volume in predicting the progression of Alzheimer’s disease because it isolates exactly where the structure is thinning.

Geometric Deep Learning: Learning the Manifold

The shift toward geometry has culminated in the use of Geometric Deep Learning. Unlike standard neural networks that look at 2D slices of an MRI, these models operate directly on 3D surface meshes.

Yunan Wu and Pierre Besson have used graph convolutional neural networks (gCNNs) to show that the complex "folded sheet" geometry of the cortex can predict fluid intelligence, age, and sex with higher accuracy than traditional scalar metrics Geometric deep learning on brain shape predicts sex and age. These models succeed because they preserve the topology of the brain—the spatial relationships between different regions—which is essential for understanding how the brain functions as a whole.

Summary: Why Geometry Works

The intuition behind this research direction is simple: Size is a consequence of shape, but shape is a reflection of process.

When a heart remodels due to cardiovascular risk Nay Aung, or a brain region atrophies due to a genetic risk factor like APOE ε4 Bolun Li, the change is rarely uniform. It is localized, directional, and topological. By moving beyond volume and embracing the rich language of differential geometry, researchers are building a more sensitive, high-resolution map of human health and disease.

Papers in this version

Showing 60 of 113 papers.

History & citationVersion 1 · 2026-10-04

Merged into another direction · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_geometric_shape_descriptors_for_high_pre_e11c2f98,
  title = {Geometric Shape Descriptors for High Precision Medical Phenotyping},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/geometric-shape-descriptors-for-high-precision-medical-phenotyping/rdr_e11c2f98b2044ec289fed10d4bd8068b},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_geometric_shape_descriptors_for_high_pre_e11c2f98_v1,
  title = {Geometric Shape Descriptors for High Precision Medical Phenotyping},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/geometric-shape-descriptors-for-high-precision-medical-phenotyping/rdr_e11c2f98b2044ec289fed10d4bd8068b/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}