FoundBioNet: A Foundation-Based Model for IDH Genotyping of Glioma from Multi-Parametric MRI

2025-08-08

AI Summary

This study addresses the clinical need for non-invasive, accurate prediction of IDH mutation status in gliomas—challenged by spatial tumor heterogeneity limiting biopsy representativeness and poor generalizability of deep learning models due to scarce annotated data. We propose a novel deep learning framework leveraging multiparametric MRI: Swin-UNeTr serves as the backbone; a Tumor-Aware Feature Encoding (TAFE) module is introduced to capture multiscale lesion characteristics; and a Cross-Modality Discrepancy (CMD) module is designed to enhance detection of the T2-FLAIR mismatch sign. A large-scale pretraining followed by task-adaptive fine-tuning strategy further improves model interpretability and cross-center robustness. Evaluated on 1,705 multicenter cases, the framework achieves AUCs of 90.58%, 88.08%, 65.41%, and 80.31% across four independent test sets—significantly outperforming baselines (p ≤ 0.05). Ablation studies confirm the efficacy of each proposed module.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Accurate, noninvasive detection of isocitrate dehydrogenase (IDH) mutation is essential for effective glioma management. Traditional methods rely on invasive tissue sampling, which may fail to capture a tumor's spatial heterogeneity. While deep learning models have shown promise in molecular profiling, their performance is often limited by scarce annotated data. In contrast, foundation deep learning models offer a more generalizable approach for glioma imaging biomarkers. We propose a Foundation-based Biomarker Network (FoundBioNet) that utilizes a SWIN-UNETR-based architecture to noninvasively predict IDH mutation status from multi-parametric MRI. Two key modules are incorporated: Tumor-Aware Feature Encoding (TAFE) for extracting multi-scale, tumor-focused features, and Cross-Modality Differential (CMD) for highlighting subtle T2-FLAIR mismatch signals associated with IDH mutation. The model was trained and validated on a diverse, multi-center cohort of 1705 glioma patients from six public datasets. Our model achieved AUCs of 90.58%, 88.08%, 65.41%, and 80.31% on independent test sets from EGD, TCGA, Ivy GAP, RHUH, and UPenn, consistently outperforming baseline approaches (p <= 0.05). Ablation studies confirmed that both the TAFE and CMD modules are essential for improving predictive accuracy. By integrating large-scale pretraining and task-specific fine-tuning, FoundBioNet enables generalizable glioma characterization. This approach enhances diagnostic accuracy and interpretability, with the potential to enable more personalized patient care.

Problem

• Noninvasive IDH genotyping for glioma using MRI
• Overcoming data scarcity in deep learning for glioma profiling
• Improving glioma biomarker accuracy with multi-modality features

Innovation

• SWIN-UNETR-based architecture for IDH prediction
• Tumor-Aware Feature Encoding for multi-scale features
• Cross-Modality Differential for T2-FLAIR mismatch