Hyperdisorder in tumor growth

TL;DR

This paper introduces the concept of hyperdisorder in tumor growth, characterized by persistent, multiscale architectural incoherence, and develops a quantitative index to analyze tumor heterogeneity and its implications for diagnosis.

Contribution

It defines hyperdisorder as a new architectural regime in tumors, introduces a multiscale quantification framework, and demonstrates its potential for improved tumor heterogeneity assessment.

Findings

Hyperdisorder persists across scales and spatial regions within tumors.

Tumor heterogeneity scales anomalously with sampling size.

Architectural disruption varies within tumor regions, affecting diagnostic sampling.

Abstract

Tumor growth is constrained by spatial, mechanical, and metabolic factors whose alignment progressively breaks down across cellular, mesoscopic, and tissue scales as tumors expand. We hypothesize that this misalignment drives tumors toward a distinct architectural regime, termed hyperdisorder. Hyperdisorder is not defined by increased heterogeneity alone, but by the coexistence of elevated disorder across scales and spatial nonstationarity within the same tumor. Unlike ordinary randomness, where independent fluctuations diminish under spatial averaging, disorder here persists, reorganizes, or even amplifies with increasing observation scale, preventing convergence toward a stable architectural description. Using hematoxylin and eosin stained whole-slide images of gastric cancer from The Cancer Genome Atlas, we quantify tumor architecture using tile-based metrics that capture complementary aspects of organization, including texture entropy, microstructural fragmentation, orientation isotropy, and multiscale entropy variation. These measures are combined into a standardized hyperdisorder index, enabling unsupervised comparison across spatial regions. We find that architectural disruption is unevenly distributed and partially decoupled across scales within individual slides, consistent with growth-driven multiscale incoherence rather than uniform stochastic variability. Testable consequences include anomalous scaling of heterogeneity with sampling size, failure of coarse graining to converge, and systematic differences between tumor cores and invasive fronts. In diagnostic and clinical contexts, this framework clarifies when measurements from limited tissue samples are representative of the whole tumor and when they are dominated by scale- and location-dependent effects.