Skip to content

bowang-lab/MedSAMSlicer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

47 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MedSAM-Lite 3D Slicer Plugin

This is the official repository for 3D Slicer Plugin for MedSAM: Segment Anything in Medical Images.

output.mp4

Installation

You can watch a video tutorial of installation steps here.

  1. Install 3D Slicer from its official website. The compatibility of our plugin has been tested with 3D Slicer >= 5.4.0
  2. Download the specific version of the plugin from releases page and extract to your desired location.
  3. In the Slicer App, press Ctrl+4 to open Extension Manager.
  4. In Install Extension tab lookup PyTorch​ and install the extension. Our plugin currently relies on this. Restart 3D Slicer afterward.
  5. Select the Welcome to Slicer drop-down menu in the toolbar at the top and navigate to Developer Tools > Extension Wizard.
  6. Click on select Extension and locate the MedSAM folder among the extracted files and directories at step 2. Accept if asked to add the new modules.
  7. Now, from the Welcome to Slicer drop-down menu, under the Segmentation sub-menu, MedSAMLite option is added. By choosing it, you get to the final steps.
  8. Choose a folder to install module dependencies and click on Install dependencies. It can take several minutes.
  9. Restart 3D Slicer.

Upgrade

Remove all pre-existing files from step#2 and install the new version as instructed before.

Usage

You can watch a video guide for usage here.

  1. From the Welcome to Slicer drop-down menu, under the Segmentation sub-menu, MedSAMLite option is added. By choosing it, you get to the final steps.
  2. Load your image file.
  3. From Prepare Data accordion menu you can choose from preset preprocessing methods (optional, but extremely helpful)
  4. From Select the Region of Interest accordion menu, you can click on Attach ROI to select the region for segmentation. It is possible to use Set Current Frame As Selection's Start and Set Current Frame As Selection's End to set ROI boundaries on the L-R slice.
  5. In Start Segmentation accordion menu, you have two different ways to infer the segmentation:
    • If you want to segment multiple regions or a single large one, it is advised to preprocess the whole image first. To do so:
      1. click on Send Image button. It will prepare the whole 3D image for the module to process. It will take several minutes. You will need to do this step only once for each image. As long as the image is not changed (new preprocessing / cropping / etc.) you do not have to redo this step.
      2. Click on Segmentation button to get the segmentation results. You can replace your ROI or delete and re-attach it (step 4) as many times as you need and repeat step 5.b without going over step 5.a again.
    • If you need to segment a single smaller region, you can bypass preprocessing and only click on Single Segmentation. After confirming, the segmentation would be inferred.
  6. In the same section, from the Segmentation and Source volume drop-down menus you can choose the right segmentation group and manually refine the segmentation results. You can also assign different color to each segmentation mask.