Related Experiment Video
Updated: Jun 30, 2025
05:45
Quantitative Immunohistochemistry of the Cellular Microenvironment in Patient Glioblastoma Resections
Published on: July 31, 2017
9.6K
Noninvasive Autopsy-Validated Tumor Probability Maps Identify Glioma Invasion Beyond Contrast Enhancement
Samuel A Bobholz1, Allison K Lowman1, Jennifer M Connelly2
1Department of Radiology, Medical College of Wisconsin, Milwaukee , Wisconsin , USA.
Neurosurgery
|March 19, 2024
Summary
This study developed radio-pathomic mapping to detect nonenhancing glioma tumor, finding it in 41.5% of patients and improving survival prediction. These maps identified infiltrative tumors beyond traditional imaging, aiding in better patient stratification.
Area of Science:
- Neuro-oncology
- Radiology
- Pathology
Background:
- Glioma often has infiltrative tumor beyond contrast-enhancing regions, impacting prognosis.
- Autopsy studies reveal significant nonenhancing tumor burden in glioma patients.
- Traditional imaging may underestimate the full extent of glioma infiltration.
Purpose of the Study:
- To identify a clinically significant subset of glioma patients with nonenhancing tumor at autopsy.
- To develop a radio-pathomic mapping method for detecting nonenhancing tumor.
- To test if autopsy-based maps can noninvasively identify infiltrative glioma beyond conventional imaging.
Main Methods:
- Retrospective study aligning 159 autopsy tissue samples from 65 subjects to nearest MRI.
- Ensemble algorithms used segmented cellularity, extracellular fluid, and cytoplasm density to predict tumor presence.
- Multi-stage models combined radiomic features (T1, T1+C, FLAIR, ADC) to generate whole-brain tumor probability maps.
Main Results:
- Tumor outside contrast enhancement was found in 41.5% of patients, associated with worse survival (HR=3.90).
- Radio-pathomic maps accurately identified nonenhancing tumor in unseen data.
- Maps detected nonenhancing tumor in 69% of presurgical cases with reduced survival (HR=1.67).
Conclusions:
- A multistage radio-pathomic model was developed using autopsy data as ground truth.
- The model successfully mapped glioma regions beyond traditional imaging signatures.
- This approach aids in identifying infiltrative tumor and improving patient stratification.
