JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

AI-driven multi-modal framework for prognostic modeling in glioblastoma: Enhancing clinical decision support

  • 0Department of Biomedical Informatics, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117597, Singapore.
Computerized Medical Imaging and Graphics : the Official Journal of the Computerized Medical Imaging Society +

|

August 15, 2025

|

August 15, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

This study introduces an AI framework using multi-modal data for glioblastoma (GBM) prognosis. The AI model integrates imaging, histopathology, and transcriptomics for improved risk stratification and personalized treatment planning.

Area Of Science

  • Oncology
  • Artificial Intelligence
  • Bioinformatics

Background

  • Glioblastoma (GBM) presents a significant challenge due to its aggressive nature and limited treatment options.
  • Current prognostic models often lack the ability to capture GBM's complex heterogeneity by relying on single data types.
  • Accurate prognostic modeling is crucial for tailoring personalized treatment strategies in GBM patients.

Purpose Of The Study

  • To develop and validate an AI-driven, multi-modal framework for enhanced clinical decision support in Glioblastoma.
  • To integrate radiological imaging, histopathology, and transcriptomic data for improved prognostic evaluation and risk stratification.
  • To utilize a Vision Transformer (ViT) for tumor grading and an attention-based deep learning model for prognostic assessment.

Main Methods

  • A Vision Transformer (ViT) was trained on FLAIR MRI scans for tumor grading (WHO grades 2-4).
  • An attention-based deep learning model integrated whole-slide histopathology images and RNA sequencing data for prognostic modeling.
  • Model performance was assessed using AUC, C-index, and Kaplan-Meier survival analysis on independent cohorts (CPTAC-GBM, TCGA-GBM).

Main Results

  • The ViT model achieved high F1-scores (>0.89) for tumor grade classification.
  • The multi-modal AI framework demonstrated superior prognostic accuracy compared to single-modality approaches, evidenced by higher C-index values.
  • Kaplan-Meier analysis showed statistically significant survival differences between high- and low-risk patient groups (p < 0.0001).

Conclusions

  • The AI-enabled multi-modal framework enhances clinical decision-making for Glioblastoma by providing accurate risk stratification.
  • Integrating radiological, histopathological, and transcriptomic data offers a comprehensive approach for personalized GBM prognosis and treatment planning.
  • This approach represents a significant advancement in leveraging AI for complex cancer data analysis and patient care.

Keywords: