Stage-Specific Benchmarking of Deep Learning Models for Glioblastoma Follow-Up MRI

2025-11-23

AI Summary

This study addresses the clinical challenge of distinguishing true tumor progression (TP) from pseudoprogression (PsP) in early follow-up MRI of glioblastoma (GBM). We propose, for the first time, a time-stage–aware deep learning benchmark framework. Using patient-level cross-validation and a quality-control–driven unified training pipeline, we systematically evaluate CNN, LSTM, Transformer, Mamba, and a hybrid Mamba+CNN architecture across multi-stage longitudinal MRI data. Results show that model discriminative performance is strongly stage-dependent, with significantly improved F1 scores and AUCs in later follow-up stages. The Mamba+CNN hybrid achieves the best trade-off among accuracy (AUC ≈ 0.74), computational efficiency, and robustness; Transformers attain comparable accuracy but at substantially higher computational cost, while lightweight CNNs are efficient yet suffer from limited generalizability. Our work demonstrates the critical importance of temporal staging in neuroimaging classification and establishes an interpretable, deployable AI paradigm for dynamic clinical risk assessment.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Differentiating true tumor progression (TP) from treatment-related pseudoprogression (PsP) in glioblastoma remains challenging, especially at early follow-up. We present the first stage-specific, cross-sectional benchmarking of deep learning models for follow-up MRI using the Burdenko GBM Progression cohort (n = 180). We analyze different post-RT scans independently to test whether architecture performance depends on time-point. Eleven representative DL families (CNNs, LSTMs, hybrids, transformers, and selective state-space models) were trained under a unified, QC-driven pipeline with patient-level cross-validation. Across both stages, accuracies were comparable (~0.70-0.74), but discrimination improved at the second follow-up, with F1 and AUC increasing for several models, indicating richer separability later in the care pathway. A Mamba+CNN hybrid consistently offered the best accuracy-efficiency trade-off, while transformer variants delivered competitive AUCs at substantially higher computational cost and lightweight CNNs were efficient but less reliable. Performance also showed sensitivity to batch size, underscoring the need for standardized training protocols. Notably, absolute discrimination remained modest overall, reflecting the intrinsic difficulty of TP vs. PsP and the dataset's size imbalance. These results establish a stage-aware benchmark and motivate future work incorporating longitudinal modeling, multi-sequence MRI, and larger multi-center cohorts.

Problem

• Differentiating true tumor progression from pseudoprogression in glioblastoma
• Benchmarking deep learning models across follow-up MRI stages
• Evaluating model performance sensitivity to time-point and architecture

Innovation

• Stage-specific deep learning benchmarking for glioblastoma MRI
• Mamba+CNN hybrid model offers best accuracy-efficiency trade-off
• Transformer variants deliver competitive AUCs with higher computational cost