🤖 AI Summary
To address the longstanding bottlenecks of low accuracy and high computational cost in automated subcortical segmentation from brain MRI, this paper proposes a novel 3D patch-based hybrid CNN-Transformer architecture. It is the first to synergistically integrate CNNs’ local texture modeling capability with Transformers’ long-range structural modeling capacity within a unified patch-processing framework, augmented by multi-scale feature fusion and patch-wise inference. Evaluated on multi-center T1-weighted MRI data, our method achieves over 10× speedup relative to FreeSurfer while improving mean Dice score by 4.2%. It significantly outperforms FastSurfer and VINN, and achieves segmentation accuracy on critical subcortical structures that approaches manual expert annotation—the clinical gold standard. The proposed approach thus unifies high accuracy, high efficiency, and full automation in subcortical segmentation.
📝 Abstract
Subcortical segmentation remains challenging despite its important applications in quantitative structural analysis of brain MRI scans. The most accurate method, manual segmentation, is highly labor intensive, so automated tools like FreeSurfer have been adopted to handle this task. However, these traditional pipelines are slow and inefficient for processing large datasets. In this study, we propose TABSurfer, a novel 3D patch-based CNN-Transformer hybrid deep learning model designed for superior subcortical segmentation compared to existing state-of-the-art tools. To evaluate, we first demonstrate TABSurfer's consistent performance across various T1w MRI datasets with significantly shorter processing times compared to FreeSurfer. Then, we validate against manual segmentations, where TABSurfer outperforms FreeSurfer based on the manual ground truth. In each test, we also establish TABSurfer's advantage over a leading deep learning benchmark, FastSurferVINN. Together, these studies highlight TABSurfer's utility as a powerful tool for fully automated subcortical segmentation with high fidelity.
