Graph-Radiomic Learning (GrRAiL) Descriptor to Characterize Imaging Heterogeneity in Confounding Tumor Pathologies

2025-09-23

AI Summary

Distinguishing benign from malignant lesions in solid tumor imaging remains challenging, primarily because conventional radiomics neglects complex spatial interdependencies among voxels of varying intensities within a lesion. To address this, we propose Graph-Radiomic Learning (GrRAiL): first, voxel-level radiomic features are clustered to construct a heterogeneity graph; second, graph-theoretic metrics are introduced to generate weighted graph descriptors that explicitly model high-order spatial dependencies. This approach breaks from traditional texture- and statistics-based radiomics by embedding graph theory—previously unexplored in radiomic frameworks—into feature representation. Evaluated on multicenter datasets, GrRAiL achieves accuracies of 78%, 74%, and 75% for glioblastoma recurrence prediction, brain metastasis classification, and pancreatic intraductal papillary mucinous neoplasm (IPMN) risk stratification, respectively—surpassing state-of-the-art methods by over 10 percentage points. The framework significantly enhances precise discrimination of histopathologically heterogeneous lesions.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Computer Vision: Medical and Biological Imaging

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

A significant challenge in solid tumors is reliably distinguishing confounding pathologies from malignant neoplasms on routine imaging. While radiomics methods seek surrogate markers of lesion heterogeneity on CT/MRI, many aggregate features across the region of interest (ROI) and miss complex spatial relationships among varying intensity compositions. We present a new Graph-Radiomic Learning (GrRAiL) descriptor for characterizing intralesional heterogeneity (ILH) on clinical MRI scans. GrRAiL (1) identifies clusters of sub-regions using per-voxel radiomic measurements, then (2) computes graph-theoretic metrics to quantify spatial associations among clusters. The resulting weighted graphs encode higher-order spatial relationships within the ROI, aiming to reliably capture ILH and disambiguate confounding pathologies from malignancy. To assess efficacy and clinical feasibility, GrRAiL was evaluated in n=947 subjects spanning three use cases: differentiating tumor recurrence from radiation effects in glioblastoma (GBM; n=106) and brain metastasis (n=233), and stratifying pancreatic intraductal papillary mucinous neoplasms (IPMNs) into no+low vs high risk (n=608). In a multi-institutional setting, GrRAiL consistently outperformed state-of-the-art baselines - Graph Neural Networks (GNNs), textural radiomics, and intensity-graph analysis. In GBM, cross-validation (CV) and test accuracies for recurrence vs pseudo-progression were 89% and 78% with>10% test-accuracy gains over comparators. In brain metastasis, CV and test accuracies for recurrence vs radiation necrosis were 84% and 74% (>13% improvement). For IPMN risk stratification, CV and test accuracies were 84% and 75%, showing>10% improvement.

Problem

• Distinguishing confounding pathologies from malignant tumors on routine imaging
• Capturing complex spatial relationships within lesion heterogeneity on MRI scans
• Improving differentiation of tumor recurrence versus radiation effects in cancers

Innovation

• Clusters sub-regions using per-voxel radiomic measurements
• Computes graph metrics to quantify spatial cluster associations
• Encodes higher-order spatial relationships via weighted graphs