AI-Based Prognostic Risk Stratification of Adult-Type Diffuse Glioma Patients from H&E-Stained Slides
7 affiliations
- 1Division of Computational Pathology, Department of Pathology and Laboratory Medicine, Indiana University School of Medicine, Indianapolis, IN 46202, USA
- 2Indiana University Melvin and Bren Simon Comprehensive Cancer Center, Indianapolis, IN 46202, USA
- 3Department of Computer Science, Kingston University, London KT1 2EE, UK
- 4Department of Neurological Surgery, Indiana University School of Medicine, Indianapolis, IN 46202, USA
- 5Department of Computer Science, Luddy School of Informatics, Computing, and Engineering, Indiana University, Indianapolis, IN 46202, USA
- 6Department of Biostatistics and Health Data Science, Indiana University School of Medicine, Indianapolis, IN 46202, USA
- 7Department of Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, IN 46202, USA
Abstract
Background/Objectives: Adult-type diffuse gliomas (ADGs) are aggressive brain tumors with highly variable clinical outcomes. Prognostic assessment directly from routine hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) could support personalized decision-making. Methods: To our knowledge, this is the first systematic benchmarking of pathology foundation models (FMs) and multiple instance learning (MIL) methods for overall survival (OS) risk stratification in ADGs under the 2021 World Health Organization (WHO) classification. A retrospective, multi-institutional cohort of 817 subjects from The Cancer Genome Atlas (TCGA) was analyzed. Following tissue segmentation and patch-level curation, we evaluated 77 pathology FM-MIL combinations (11 FMs × 7 MIL methods), together with seven ResNet-50 baseline configurations (84 encoder-MIL configurations in total), using patient-level 10-fold cross-validation. A perturbation-based evaluation assessed attention mechanism faithfulness. Results: UNI2 paired with MambaMIL achieved the highest cross-validated concordance index (C-index = 0.77), whereas its time-dependent AUC was modest (0.60–0.61 at 12–36 months), and several combinations performed comparably. Kaplan–Meier analysis demonstrated significant separation between high- and low-risk groups (log-rank p = 9.16 × 10−17; hazard ratio [HR] = 2.90, 95% confidence interval [CI]: 2.23–3.76), with a median OS of 16.6 months (95% CI: 15.0–20.4) versus 63.5 months (95% CI: 43.9-NR), respectively. Tertile stratification confirmed a monotonic OS gradient from 94.5 months to 14.7 months (p = 3.25 × 10−17). Risk stratification within individual WHO 2021 types was not significant. Most domain-specific FMs outperformed the ImageNet-pretrained baseline, and the removal of highly attended patches reduced performance, supporting attention faithfulness. Conclusions: The AI-derived risk score remained associated with OS after adjustment for individual clinical and molecular variables, but independent prognostic value was not demonstrated in the fully adjusted model (HR = 1.32, 95% CI: 0.94–1.85, p = 0.104), and its discrimination largely reflects differences between WHO 2021 types. This benchmarking framework provides guidance for FM-MIL design in computational pathology applications targeting survival prediction; external validation is required before clinical application.
Related
- Prognostic Value of Systemic Inflammation Markers and Clinical Factors in Grade 4 Diffuse Glioma: A Multicenter Retrospective Real-World Study.
- Implementing blended cognitive behavioral therapy for severe fatigue in routine hospital care for diffuse glioma patients: a qualitative study
- Validation and prognostic utility of the mSMART 4.0 risk stratification system in Chinese patients with multiple myeloma
- Reclassifying renal cell carcinoma with pathologic lymph node metastasis: An approach to refining prognostic stratification in patients with AJCC III and IV stages
- Adult Thalamic Diffuse Midline Glioma, H3 K27-Altered, With KRAS Comutation and Potential Early Subependymal Dissemination
BibTeX
@article{innani2026aibased,
title = {{AI-Based Prognostic Risk Stratification of Adult-Type Diffuse Glioma Patients from H&E-Stained Slides}},
author = {Shubham Innani and Carla Pitarch-Abaigar and Marwan M. Majeed and Hannah Harmsen and W. Robert Bell and Dimitrios Makris and Spyridon Bakas},
journal = {Cancers},
year = {2026},
month = oct,
doi = {10.3390/cancers18203259},
url = {https://www.mdpi.com/2072-6694/18/20/3259},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - AI-Based Prognostic Risk Stratification of Adult-Type Diffuse Glioma Patients from H&E-Stained Slides
AU - Shubham Innani
AU - Carla Pitarch-Abaigar
AU - Marwan M. Majeed
AU - Hannah Harmsen
AU - W. Robert Bell
AU - Dimitrios Makris
AU - Spyridon Bakas
PY - 2026
DA - 2026/10/01
JO - Cancers
DO - 10.3390/cancers18203259
UR - https://www.mdpi.com/2072-6694/18/20/3259
AB - Background/Objectives: Adult-type diffuse gliomas (ADGs) are aggressive brain tumors with highly variable clinical outcomes. Prognostic assessment directly from routine hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) could support personalized decision-making. Methods: To our knowledge, this is the first systematic benchmarking of pathology foundation models (FMs) and multiple instance learning (MIL) methods for overall survival (OS) risk stratification in ADGs under the 2021 World Health Organization (WHO) classification. A retrospective, multi-institutional cohort of 817 subjects from The Cancer Genome Atlas (TCGA) was analyzed. Following tissue segmentation and patch-level curation, we evaluated 77 pathology FM-MIL combinations (11 FMs × 7 MIL methods), together with seven ResNet-50 baseline configurations (84 encoder-MIL configurations in total), using patient-level 10-fold cross-validation. A perturbation-based evaluation assessed attention mechanism faithfulness. Results: UNI2 paired with MambaMIL achieved the highest cross-validated concordance index (C-index = 0.77), whereas its time-dependent AUC was modest (0.60–0.61 at 12–36 months), and several combinations performed comparably. Kaplan–Meier analysis demonstrated significant separation between high- and low-risk groups (log-rank p = 9.16 × 10−17; hazard ratio [HR] = 2.90, 95% confidence interval [CI]: 2.23–3.76), with a median OS of 16.6 months (95% CI: 15.0–20.4) versus 63.5 months (95% CI: 43.9-NR), respectively. Tertile stratification confirmed a monotonic OS gradient from 94.5 months to 14.7 months (p = 3.25 × 10−17). Risk stratification within individual WHO 2021 types was not significant. Most domain-specific FMs outperformed the ImageNet-pretrained baseline, and the removal of highly attended patches reduced performance, supporting attention faithfulness. Conclusions: The AI-derived risk score remained associated with OS after adjustment for individual clinical and molecular variables, but independent prognostic value was not demonstrated in the fully adjusted model (HR = 1.32, 95% CI: 0.94–1.85, p = 0.104), and its discrimination largely reflects differences between WHO 2021 types. This benchmarking framework provides guidance for FM-MIL design in computational pathology applications targeting survival prediction; external validation is required before clinical application.
SN - 2072-6694
ER -