Segmentation of Gray Matters and White Matters from Brain MRI data

2026-03-30

AI Summary

This study addresses the poor generalizability and task-specific tuning requirements of existing brain MRI gray matter (GM) and white matter (WM) segmentation methods. It proposes the first adaptation of the MedSAM foundation model into a three-class (background, GM, WM) segmentation framework by minimally modifying the architecture—extending only the mask decoder while fine-tuning the prompt encoder and decoder, and freezing the image encoder to preserve its general-purpose representations. Leveraging FSL BET for skull stripping and FAST tissue probability maps as preprocessing, the method trains on 2D multi-view slices from the IXI dataset, enabling efficient inference. With minimal architectural changes, the approach achieves substantial gains in segmentation accuracy, attaining Dice scores of up to 0.8751 for both GM and WM, thereby demonstrating the strong transferability of medical foundation models to multi-tissue segmentation tasks.

Abstract

Accurate segmentation of brain tissues such as gray matter and white matter from magnetic resonance imaging is essential for studying brain anatomy, diagnosing neurological disorders, and monitoring disease progression. Traditional methods, such as FSL FAST, produce tissue probability maps but often require task-specific adjustments and face challenges with diverse imaging conditions. Recent foundation models, such as MedSAM, offer a prompt-based approach that leverages large-scale pretraining. In this paper, we propose a modified MedSAM model designed for multi-class brain tissue segmentation. Our preprocessing pipeline includes skull stripping with FSL BET, tissue probability mapping with FSL FAST, and converting these into 2D axial, sagittal, coronal slices with multi-class labels (background, gray matter, and white matter). We extend MedSAM's mask decoder to three classes, freezing the pre-trained image encoder and fine-tuning the prompt encoder and decoder. Experiments on the IXI dataset achieve Dice scores up to 0.8751. This work demonstrates that foundation models like MedSAM can be adapted for multi-class medical image segmentation with minimal architectural modifications. Our findings suggest that such models can be extended to more diverse medical imaging scenarios in future work.

Problem

• brain MRI
• gray matter
• white matter
• tissue segmentation
• medical image analysis

Innovation

• MedSAM
• multi-class segmentation
• brain MRI
• foundation model adaptation
• prompt-based segmentation