Dr. Mohit Agarwal – Neuroradiology at Medical College of Wisconsin

Overview

A Comparison of Radio-Pathomic, Diffusion, and Perfusion Imaging Features for Identifying Pseudoprogression in Glioblastoma.

Aug 7, 2026DOI: 10.3174/ajnr.a9574Journal ArticleAJNR

Written by: Hope M ReecherSavannah R DuenwegAllison K LowmanAleksandra WiniarzBiprojit NathBenjamin ChaoDaniel KimMichael J BarrettFitzgerald KyeremeJennifer ConnellyElaine TanhehcoMax O KrucoffRupen DesaiMohit AgarwalJamie JacobsohnJennifer TuscherPeter S LaVioletteSamuel A Bobholz

Find these authors on Roon

Dr. Mohit Agarwal

Neuroradiology · MCW

Summary

INTRODUCTION: Glioblastoma is an aggressive brain tumor with poor survival rates and nearly universal recurrence. Magnetic resonance (MR) imaging obtained after radiation and chemotherapy treatment often shows "pseudoprogression" (PsP), or contrast enhancement that mimics tumor growth. However, this does not indicate actual disease progression. This study compared the efficacy of a previously developed radio-pathomic mapping model for discriminating between true progression (TP) and PsP relative to advanced diffusion and perfusion metrics.MATERIALS AND METHODS: This retrospective study utilized the UCSD-PTGBM single-site public dataset comprised of 171 MR imaging sessions (139 TP and 32 PsP) from 130 adult patients following initial biopsy or resection. Radio-pathomic maps of cell density (RPM Cell) and extracellular fluid density (RPM ECF) from a previously-established methodology were created using conventional imaging data (e.g., T1, T1C, FLAIR, ADC) from the public UCSD dataset. These maps were then compared to perfusion (rCBV, CBF, MTT) and restricted spectrum imaging (RSI) diffusion (Free, Hindered, Cell) features to assess differences between TP and PsP within regions of contrast enhancement. Statistical analyses involved Pearson correlation coefficients, mixed effect models, and receiver operator characteristic (ROC) analysis with area under the curve (AUC) used to evaluate performance at the first and all timepoints, using DeLong's tests to confirm significance.RESULTS: RPM Cell significantly (p=0.006) outperformed other imaging features in discriminating TP from PsP (AUC=0.827 vs. <0.687 for alternatives, across all timepoints). Features that differed between patients with TP and PsP included RPM Cell (B=5.71, p<0.001) and RSI Cell (B=2.54, p=0.01).CONCLUSION: RPM Cell shows promise as a novel, accessible marker for detecting tumor recurrence. These cell density maps outperform current advanced imaging techniques in distinguishing true tumor progression from PsP in a retrospective single-site dataset of glioblastoma patients, potentially offering clinicians valuable insights into the tumor microenvironment through a non-invasive modality without requiring additional contrast agents or scan time.