AI-Driven MRI-based Brain Tumour Segmentation Benchmarking

TL;DR

This paper evaluates various promptable AI models for MRI brain tumor segmentation on the BraTS dataset, comparing their zero-shot and fine-tuned performances to establish their strengths and limitations.

Contribution

It provides a comprehensive benchmarking of promptable models like SAM and MedSAM against nnU-Net on medical datasets, highlighting their capabilities and challenges in medical image segmentation.

Findings

SAM and SAM 2 achieve high Dice scores with accurate bounding box prompts.

nnU-Net remains the most practical and effective model for medical segmentation.

Fine-tuning improves prompt-based performance but does not surpass nnU-Net.

Abstract

Medical image segmentation has greatly aided medical diagnosis, with U-Net based architectures and nnU-Net providing state-of-the-art performance. There have been numerous general promptable models and medical variations introduced in recent years, but there is currently a lack of evaluation and comparison of these models across a variety of prompt qualities on a common medical dataset. This research uses Segment Anything Model (SAM), Segment Anything Model 2 (SAM 2), MedSAM, SAM-Med-3D, and nnU-Net to obtain zero-shot inference on the BraTS 2023 adult glioma and pediatrics dataset across multiple prompt qualities for both points and bounding boxes. Several of these models exhibit promising Dice scores, particularly SAM and SAM 2 achieving scores of up to 0.894 and 0.893, respectively when given extremely accurate bounding box prompts which exceeds nnU-Net's segmentation performance. However, nnU-Net remains the dominant medical image segmentation network due to the impracticality of providing highly accurate prompts to the models. The model and prompt evaluation, as well as the comparison, are extended through fine-tuning SAM, SAM 2, MedSAM, and SAM-Med-3D on the pediatrics dataset. The improvements in point prompt performance after fine-tuning are substantial and show promise for future investigation, but are unable to achieve better segmentation than bounding boxes or nnU-Net.