Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TABSurfer: a Hybrid Deep Learning Architecture for Subcortical Segmentation

Aaron Cao, Vishwanatha M. Rao, Kejia Liu, Xinrui Liu, Andrew F. Laine, Jia Guo

TL;DR

TABSurfer addresses the challenge of automated subcortical segmentation by integrating a 3D patch-based CNN with a Vision Transformer bridge to capture long-range context while preserving local detail. Trained and evaluated on a large, heterogeneous T1-weighted MRI dataset, it outperforms FreeSurfer and the FastSurferVINN benchmark against both automated and manual ground truths, and it delivers fast whole-brain segmentation per patch-based reconstruction. The approach demonstrates robust cross-dataset performance and smoother, more accurate contours, signaling a valuable direction for scalable, high-fidelity subcortical analysis in large-scale studies. While the method is computationally intensive and slower than some competitors, its accuracy gains motivate further work on efficiency and broader generalizability.

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.

TABSurfer: a Hybrid Deep Learning Architecture for Subcortical Segmentation

TL;DR

TABSurfer addresses the challenge of automated subcortical segmentation by integrating a 3D patch-based CNN with a Vision Transformer bridge to capture long-range context while preserving local detail. Trained and evaluated on a large, heterogeneous T1-weighted MRI dataset, it outperforms FreeSurfer and the FastSurferVINN benchmark against both automated and manual ground truths, and it delivers fast whole-brain segmentation per patch-based reconstruction. The approach demonstrates robust cross-dataset performance and smoother, more accurate contours, signaling a valuable direction for scalable, high-fidelity subcortical analysis in large-scale studies. While the method is computationally intensive and slower than some competitors, its accuracy gains motivate further work on efficiency and broader generalizability.

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.
Paper Structure (12 sections, 3 figures, 2 tables)

This paper contains 12 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Data
  4. Pipeline and Model Architecture
  5. Model Training
  6. Model Evaluation
  7. Results
  8. Evaluation on FreeSurfer Segmentations
  9. Evaluation on Manual Segmentations
  10. Discussion and Conclusion
  11. Compliance with ethical standards
  12. Acknowledgments

Figures (3)

  • Figure 1: Age and Gender Distributions for each Dataset
  • Figure 2: a. Our pipeline extracts 3D patches from the input scan, feeds them into our model, and reconstructs the output predicted classes to generate a segmentation. b. Visualization of the TABSurfer model's architecture.
  • Figure 3: Sample predicted slices and volumes between FreeSurfer, TABSurfer, and FastSurferVINN segmentations