Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Robust Computational Extraction of Non-Enhancing Hypercellular Tumor Regions from Clinical Imaging Data

A. Brawanski, Th. Schaffer, F. Raab, K. -M. Schebesch, M. Schrey, Chr. Doenitz, A. M. Tomé, E. W. Lang

TL;DR

This study presents a robust, multi-architecture framework to produce probability maps of non-enhancing hypercellular (NEH) tumor regions from routine MRI, leveraging PAUNet and UNETR++ to capture both local and global spatial features. The NEH maps are smoothed and thresholded to yield clinically interpretable segments, with a dual-model consensus approach and a supplementary STAPLE-like fusion. Validation against independent markers—ap-rCBV and first-recurrence ET—via permutation testing demonstrates that NEH regions exhibit biologically plausible hyperperfusion and spatial proximity to recurrence, supporting their relevance as imaging biomarkers for precision neuro-oncology. The framework is designed to inform surgical planning and targeted therapy while highlighting the need for histopathologic correlation and prospective, multi-institution validation to enable clinical adoption.

Abstract

Accurate identification of non-enhancing hypercellular (NEH) tumor regions is an unmet need in neuro-oncological imaging, with significant implications for patient management and treatment planning. We present a robust computational framework that generates probability maps of NEH regions from routine MRI data, leveraging multiple network architectures to address the inherent variability and lack of clear imaging boundaries. Our approach was validated against independent clinical markers -- relative cerebral blood volume (rCBV) and enhancing tumor recurrence location (ETRL) -- demonstrating both methodological robustness and biological relevance. This framework enables reliable, non-invasive mapping of NEH tumor compartments, supporting their integration as imaging biomarkers in clinical workflows and advancing precision oncology for brain tumor patients.

Robust Computational Extraction of Non-Enhancing Hypercellular Tumor Regions from Clinical Imaging Data

TL;DR

This study presents a robust, multi-architecture framework to produce probability maps of non-enhancing hypercellular (NEH) tumor regions from routine MRI, leveraging PAUNet and UNETR++ to capture both local and global spatial features. The NEH maps are smoothed and thresholded to yield clinically interpretable segments, with a dual-model consensus approach and a supplementary STAPLE-like fusion. Validation against independent markers—ap-rCBV and first-recurrence ET—via permutation testing demonstrates that NEH regions exhibit biologically plausible hyperperfusion and spatial proximity to recurrence, supporting their relevance as imaging biomarkers for precision neuro-oncology. The framework is designed to inform surgical planning and targeted therapy while highlighting the need for histopathologic correlation and prospective, multi-institution validation to enable clinical adoption.

Abstract

Accurate identification of non-enhancing hypercellular (NEH) tumor regions is an unmet need in neuro-oncological imaging, with significant implications for patient management and treatment planning. We present a robust computational framework that generates probability maps of NEH regions from routine MRI data, leveraging multiple network architectures to address the inherent variability and lack of clear imaging boundaries. Our approach was validated against independent clinical markers -- relative cerebral blood volume (rCBV) and enhancing tumor recurrence location (ETRL) -- demonstrating both methodological robustness and biological relevance. This framework enables reliable, non-invasive mapping of NEH tumor compartments, supporting their integration as imaging biomarkers in clinical workflows and advancing precision oncology for brain tumor patients.
Paper Structure (19 sections, 4 figures, 7 tables)

This paper contains 19 sections, 4 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Recent related own work
  3. Methods
  4. Extraction and estimation of NEH regions
  5. Clinical validation of segmentation models
  6. Testing H1 and H2 with the UPenn-GBM dataset
  7. Testing H3 using the RHUH glioblastoma dataset
  8. Evaluation metrics
  9. Results
  10. Quality assesment and feasibility of NEH segmentation
  11. Clinical validation of segmentation methods
  12. Comparison of rCBV values
  13. Comparison of preoperative NEH to first recurrence
  14. Discussion
  15. Biological interpretation
  16. ...and 4 more sections

Figures (4)

  • Figure 1: Subfigure 1a shows a plain T1 with contrast, 1b the PAUNET segmentation, 1c the UNETR++ segmentation, 1d UNETR-staple++ based segmentation. In the images with 4 segments ( 1b-c): green: edema, yellow: contrast enhancement, red: necrosis, blue: NEH. In 1d green denotes low probability, red denotes high probability
  • Figure 2: Mean CBV intensities by PAUNet and UNETR++
  • Figure 3: Intensities of the different masked rCBV areas (left subimage). The segmented NEH segment is significantly higher than all the other inflated areas (right subimage)
  • Figure 4: Preoperative NEH segment (red) warped on the recurrent ET compartment