DRBD-Mamba for Robust and Efficient Brain Tumor Segmentation with Analytical Insights

2025-10-16

AI Summary

Brain tumor segmentation faces dual challenges of accuracy and robustness due to subregion heterogeneity; existing Mamba-based models suffer from high computational overhead caused by multi-axis sequential modeling and poor generalization across diverse data distributions. To address these issues, we propose Dual-Resolution Bidirectional Mamba (DRBD-Mamba): (1) it employs space-filling curves for efficient 3D-to-1D feature mapping; (2) it introduces a dual-resolution Mamba backbone, reversible gated contextual fusion, and feature quantization to jointly enhance representational capacity and robustness; and (3) it establishes five systematic BraTS data splits for rigorous and reproducible evaluation. On the BraTS2023 test set, DRBD-Mamba achieves absolute Dice score improvements of +1.75% for tumor core and +0.93% for enhancing tumor, with average gains of +0.86% and +1.45% under a novel five-fold cross-validation protocol. Moreover, inference speed is accelerated by 15× over state-of-the-art methods, demonstrating superior efficiency and performance.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Ensemble Methods
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Accurate brain tumor segmentation is significant for clinical diagnosis and treatment. It is challenging due to the heterogeneity of tumor subregions. Mamba-based State Space Models have demonstrated promising performance. However, they incur significant computational overhead due to sequential feature computation across multiple spatial axes. Moreover, their robustness across diverse BraTS data partitions remains largely unexplored, leaving a critical gap in reliable evaluation. To address these limitations, we propose dual-resolution bi-directional Mamba (DRBD-Mamba), an efficient 3D segmentation model that captures multi-scale long-range dependencies with minimal computational overhead. We leverage a space-filling curve to preserve spatial locality during 3D-to-1D feature mapping, thereby reducing reliance on computationally expensive multi-axial feature scans. To enrich feature representation, we propose a gated fusion module that adaptively integrates forward and reverse contexts, along with a quantization block that discretizes features to improve robustness. In addition, we propose five systematic folds on BraTS2023 for rigorous evaluation of segmentation techniques under diverse conditions and present detailed analysis of common failure scenarios. On the 20% test set used by recent methods, our model achieves Dice improvements of 0.10% for whole tumor, 1.75% for tumor core, and 0.93% for enhancing tumor. Evaluations on the proposed systematic five folds demonstrate that our model maintains competitive whole tumor accuracy while achieving clear average Dice gains of 0.86% for tumor core and 1.45% for enhancing tumor over existing state-of-the-art. Furthermore, our model attains 15 times improvement in efficiency while maintaining high segmentation accuracy, highlighting its robustness and computational advantage over existing approaches.

Problem

• Addressing computational inefficiency in brain tumor segmentation models
• Improving robustness across diverse brain tumor data partitions
• Capturing multi-scale long-range dependencies with minimal overhead

Innovation

• Dual-resolution bi-directional Mamba captures multi-scale dependencies
• Space-filling curve preserves locality in 3D-to-1D mapping
• Gated fusion module adaptively integrates forward and reverse contexts