In this repo I try to collect a list of papers and projects that I find interesting that utilizes the Segment Anything Model (SAM) to perform segmentation on medical images, utilizes SAM as part of the frameworks, or perform anaylsis or studies of SAM on medical images. Note that this SAM is a relative new model and there may be ignored papers or works that are ignored.
For a list of all projects and researches of SAM in various fields, check out the Awesome Segment Anything repository.
If you find any interesting works feel free to create pull requests or email me to make the list more comprehensive.
| Title | Paper | Code | Dataset | Keywords | Comments |
|---|---|---|---|---|---|
| SAM.MD: Zero-shot medical image segmentation capabilities of the Segment Anything Mode | arxiv | - | abdominal CT organ | zero-shot | comparasion between prompted SAM and 2D and 3D nnUNet |
| Segment Anything Model (SAM) for Digital Pathology: Assess Zero-shot Segmentation on Whole Slide Imaging | arxiv | - | Skin tumor, Skin tissue | zero-shot | comparasion between prompted SAM and SimTriplet |
| When SAM Meets Medical Images: An Investigation of Segment Anything Model (SAM) on Multi-phase Liver Tumor Segmentation | arxiv | - | Liver Tumor | zero-shot | comparasion between SAM and UNet |
| SAM vs BET: A Comparative Study for Brain Extraction and Segmentation of Magnetic Resonance Images using Deep Learning | arxiv | - | Brain MRI | zero-shot | brain extraction |
| The “Segment Anything” foundation model achieves favorable brain tumor autosegmentation accuracy on MRI to support radiotherapy treatment planning | arxiv | - | Brain MRI | using BraTS 2020 dataset, use promting for SAM | |
| Accuracy of Segment-Anything Model (SAM) in Medical Image Segmentation Tasks | arxiv | - | Various | Datasets: ACDC, LiTS, Hipp, ISIC, Prostate, LA, BraTS, Pancreas, BUID, Kvasir, CIR. Benchmark Models: U-Net, U-Net++, Attention U-Net, Trans U-Net, UCTransNet, SAM, SAM-Points, SAM-Boxes (3 prompting settings). | |
| Can SAM Segment Polyps? | arxiv | - | Polyp | Utilizes unprompted settings for SAM. Compute S-measure (Sα) score values for the N masks, and the mask with the highest score is selected as the segmentation map. | |
| Segment Anything Model for Medical Image Analysis: an Experimental Study | arxiv | Code | Various | Tests SAM on 11 medical datasets. Compares results with RITM. Experiments with various number of prompts. | |
| Segment Anything in Medical Images | arxiv | Code | Various | Development of fine-tuning method to adapt SAM to general medical image segmentation. | |
| Medical SAM Adapter: Adapting Segment Anything Model for Medical Image Segmentation | arxiv | Code | Various | Introduction of Medical SAM Adapter (MSA), a bottle-neck model to fine-tune the SAM model. Various datasets (AMOS2022, BTCV, and etc) are used |
| Title | Demo | Paper | Code | Comments |
|---|---|---|---|---|
| SAMM (Segment Any Medical Model): A 3D Slicer Integration to SAM |  |
arxiv | code | SAM integration in 3D Slicer for semi-automatic segmentation |
| Segment Anything Model (SAM) in Napari |  |
- | code | SAM integration in 3D Napari for click-based semantic segmentation |
| SAM Medical Imaging |  |
- | code | SAM segmentaiton of DICOM files using Colab |
| Segment-Anything-Automatically-on-Medical-Image (SAAMI) |  |
- | code | Automatic SAM 3D segmentation mask generation without prompting |