Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SAM vs BET: A Comparative Study for Brain Extraction and Segmentation of Magnetic Resonance Images using Deep Learning

Sovesh Mohapatra, Advait Gosai, Gottfried Schlaug

TL;DR

The study directly compares the Segment Anything Model (SAM) against FSL BET for automatic brain extraction and within-brain segmentation across diverse MRI datasets, including lesions. By converting 3D MRIs to 2D slices and applying a bounding-box-based SAM workflow, the authors demonstrate that SAM generally achieves higher Dice, IoU, and accuracy than BET, especially in low-quality images and near outer-brain regions. A standout result on a WMH 3DT1 dataset shows SAM markedly outperforming BET (Dice 0.914 vs 0.628; IoU 0.842 vs 0.518; Accuracy 0.871 vs 0.642; Recall 0.944 vs 0.306). The findings suggest SAM’s potential as a robust, versatile tool for brain extraction and targeted structure/lesion segmentation, with caveats related to computational demands and 2D-to-3D reconstruction nuances.

Abstract

Brain extraction is a critical preprocessing step in various neuroimaging studies, particularly enabling accurate separation of brain from non-brain tissue and segmentation of relevant within-brain tissue compartments and structures using Magnetic Resonance Imaging (MRI) data. FSL's Brain Extraction Tool (BET), although considered the current gold standard for automatic brain extraction, presents limitations and can lead to errors such as over-extraction in brains with lesions affecting the outer parts of the brain, inaccurate differentiation between brain tissue and surrounding meninges, and susceptibility to image quality issues. Recent advances in computer vision research have led to the development of the Segment Anything Model (SAM) by Meta AI, which has demonstrated remarkable potential in zero-shot segmentation of objects in real-world scenarios. In the current paper, we present a comparative analysis of brain extraction techniques comparing SAM with a widely used and current gold standard technique called BET on a variety of brain scans with varying image qualities, MR sequences, and brain lesions affecting different brain regions. We find that SAM outperforms BET based on average Dice coefficient, IoU and accuracy metrics, particularly in cases where image quality is compromised by signal inhomogeneities, non-isotropic voxel resolutions, or the presence of brain lesions that are located near (or involve) the outer regions of the brain and the meninges. In addition, SAM has also unsurpassed segmentation properties allowing a fine grain separation of different issue compartments and different brain structures. These results suggest that SAM has the potential to emerge as a more accurate, robust and versatile tool for a broad range of brain extraction and segmentation applications.

SAM vs BET: A Comparative Study for Brain Extraction and Segmentation of Magnetic Resonance Images using Deep Learning

TL;DR

The study directly compares the Segment Anything Model (SAM) against FSL BET for automatic brain extraction and within-brain segmentation across diverse MRI datasets, including lesions. By converting 3D MRIs to 2D slices and applying a bounding-box-based SAM workflow, the authors demonstrate that SAM generally achieves higher Dice, IoU, and accuracy than BET, especially in low-quality images and near outer-brain regions. A standout result on a WMH 3DT1 dataset shows SAM markedly outperforming BET (Dice 0.914 vs 0.628; IoU 0.842 vs 0.518; Accuracy 0.871 vs 0.642; Recall 0.944 vs 0.306). The findings suggest SAM’s potential as a robust, versatile tool for brain extraction and targeted structure/lesion segmentation, with caveats related to computational demands and 2D-to-3D reconstruction nuances.

Abstract

Brain extraction is a critical preprocessing step in various neuroimaging studies, particularly enabling accurate separation of brain from non-brain tissue and segmentation of relevant within-brain tissue compartments and structures using Magnetic Resonance Imaging (MRI) data. FSL's Brain Extraction Tool (BET), although considered the current gold standard for automatic brain extraction, presents limitations and can lead to errors such as over-extraction in brains with lesions affecting the outer parts of the brain, inaccurate differentiation between brain tissue and surrounding meninges, and susceptibility to image quality issues. Recent advances in computer vision research have led to the development of the Segment Anything Model (SAM) by Meta AI, which has demonstrated remarkable potential in zero-shot segmentation of objects in real-world scenarios. In the current paper, we present a comparative analysis of brain extraction techniques comparing SAM with a widely used and current gold standard technique called BET on a variety of brain scans with varying image qualities, MR sequences, and brain lesions affecting different brain regions. We find that SAM outperforms BET based on average Dice coefficient, IoU and accuracy metrics, particularly in cases where image quality is compromised by signal inhomogeneities, non-isotropic voxel resolutions, or the presence of brain lesions that are located near (or involve) the outer regions of the brain and the meninges. In addition, SAM has also unsurpassed segmentation properties allowing a fine grain separation of different issue compartments and different brain structures. These results suggest that SAM has the potential to emerge as a more accurate, robust and versatile tool for a broad range of brain extraction and segmentation applications.
Paper Structure (16 sections, 4 figures, 1 table)

This paper contains 16 sections, 4 figures, 1 table.

Table of Contents

  1. Main
  2. Methodology
  3. Selection of MRI data
  4. Preprocessing of MRI data
  5. Brain Extraction using BET
  6. Brain Extraction using SAM
  7. 3D MRI scan preprocessing and reconstruction
  8. Evaluation metrics for brain extraction
  9. Results
  10. Evaluation of Extractions by BET and SAM
  11. Visual Comparison
  12. Applications of SAM segmenting particular brain structures and lesions
  13. Discussion
  14. Advantages and Limitations of SAM and BET
  15. Future Work
  16. ...and 1 more sections

Figures (4)

  • Figure 1: Comprehensive SAM workflow – visualizing the end-to-end process with custom box algorithm, highlighting blue dots (exclusion mask) and yellow dots (inclusion mask)
  • Figure 2: Comparison of original MRI images and extraction outputs by BET and SAM for A. sagittal, B. coronal, and C. axial anatomical planes. BET extraction results in sagittal, coronal, and axial planes, respectively. SAM extraction results in sagittal, coronal, and axial planes, respectively. The data is from ATLAS v2.0 dataset liew2018large.
  • Figure 3: Comparison of original MRI images and extraction outputs by BET and SAM for A. T2 scan from a chronic stroke patient in the coronal plane, B. FLAIR scan from a chronic stroke patient in the sagittal plane, C. 3DT1 scan from a patient having WMH in the coronal plane, and D. FA scan from a patient having chronic stroke in the coronal plane. Data for A., B. and D. are from in-house dataset, and C. is from WMH challenge dataset kuijf2019standardized.
  • Figure 4: A., B. Ventricles extracted from a T1 scan from acute stroke phase. C. Corpus callosum extracted from a 3DT1 scan from chronic stroke phase. D. Lesions extracted from a DWI scan from chronic stroke phase. E. Microbleeds extracted from a SWAN scan from chronic stroke phase. Data in A. and B. are from ATLAS v2.0 liew2018large, and C., D., and E. are from in-house dataset.