From Anatomy to Outcome: Linking Glioma Location Patterns to Survival Using Non-Negative Matrix Factorization

Read full text Summarise with VIRAT
Authors: Marianne Schell, Joel Kohler, Martha Foltyn‐Dumitru, Haidar Alzaid, Irada Pflüger, Katharina Schregel, Tim Hilgenfeld, Julius M. Kernbach, Sandro Manuel Krieg, Felix Sahm, Wolfgang Wick, Martin Bendszus
+1 more Jessica Jesser
Cited by: 0
References: 42

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

PURPOSE: The anatomical distribution of IDH-wildtype gliomas may encode prognostic information not captured by conventional clinical or molecular markers. The purpose of this study was to determine whether continuous, data-driven spatial patterns of tumor involvement provide prognostic value beyond established clinical and molecular factors. METHODS: We applied non-negative matrix factorization (NMF) to preoperative tumor segmentations from 429 patients with IDH-wildtype gliomas to identify coherent spatial components of tumor involvement. Six reproducible spatial signatures were extracted and integrated into multivariate Cox proportional hazards models. Models additionally included tumor volume, age, sex, O6-methylguanine-DNA methyltransferase (MGMT) promotor methylation status, Eastern Cooperative Oncology Group (ECOG) performance status, and extent of resection. RESULTS: All six spatial patterns demonstrated significant association with overall survival (p ≤ 0.005). Notably, involvement of the left parietal-occipital regions was linked to the poorest outcomes, while temporal patterns showed weaker associations with survival. Incorporating spatial patterns into the prognostic model improved its discriminatory power and altered the influence of tumor volume, indicating an interaction between tumor location and tumor burden. CONCLUSION: Continuous spatial phenotyping of IDH-wildtype gliomas using NMF captures prognostically relevant information not reflected by conventional markers. Integration of spatial patterns into prognostic models enhances risk stratification and supports the incorporation of detailed spatial analyses into radiologic workflows.

Fields of Study & Subjects

Glioma Diagnosis and Treatment Radiomics and Machine Learning in Medical Imaging Brain Tumor Detection and Classification

Publication details

Journal / source
Clinical Neuroradiology
Volume
36
Issue
3
Pages
921–930
Publication date
27 March 2026
Language
EN