Skip to content

mohitulm/InsitumProject_VesselFMnnInteractiveMITK

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 

Repository files navigation

VesselFM MITK Plugin

Integration of VesselFM into MITK Workbench as a native plugin and segmentation tool.

This project enables running deep-learning-based vessel segmentation directly inside MITK, including:

  • a Workbench view plugin for standalone inference
  • a Segmentation Tool (Preview Tool) for interactive workflows

Features

  • Run VesselFM directly inside MITK
  • Export selected images automatically to NIfTI
  • Execute Python-based inference via Hydra
  • Load predicted segmentation masks back into MITK
  • Interactive segmentation via Preview/Confirm workflow
  • Configurable:
    • device (CPU / CUDA)
    • threshold for probabiltiy of counting into the segmentated pixel-> low (i.e. 0.2) more pixels will count into the final segmentation (could lead to more wrong segmented pixels but also maybe a more detailed vessel mask; medium (0.5) default option, balance; high (i.e. 0.8) less pixels are segmentated (potentially fewer bone missegmentation but also more sparse vessel mask)
    • restrict segmentation around existing roi with padding
    • compute volume and pixel spacing information
    • checkpoint path -> useful i.e. if you did pretraining
    • Get volume metrics -> after segmentation is done you can get additional volumetric information

📁 Repository Structure

Plugins/org.mitk.gui.qt.vesselfm/ QmitkVesselFMView.cpp VesselFMRunner.cpp mitkVesselFMActivator.cpp

Modules/VesselFM/ Interactions/mitkVesselFMSegTool3D.cpp -> main lgic file (optional GUI files)

⚙️ Requirements

  • System: Windows 10/11
  • Visual Studio 2022 (MSVC)
  • CMake ≥ 3.20

Dependencies

  • MITK source (e.g. 2025.08 / 2025.12)
  • Qt 6 (MSVC build!)
  • Python environment with:
  • torch
  • hydra-core
  • monai
  • VesselFM dependencies
  1. MITK Superbuild

MITK must be built from source to ensure plugin compatibility -> This is a multi step process further information please check put the official build instructions from mitk: https://docs.mitk.org/nightly/BuildInstructionsPage.html

Example Directory layout: C:\MITK-2025.12\ <- MITK source C:\MITK_build\MITK-build\ <- build directory

  1. After you built the MITK superbuild you mith have to built the workbench (again) if there is no .exe file yet -> run this command for this: cmake --build C:\MITK_build\MITK-build --config Release --target MitkWorkbench

  2. If you make changes to the plugin rebuild only the change files i.e. with these commands

    • make --build C:\MITK_build\MITK-build --config Release --target org_mitk_gui_qt_vesselfm
    • cmake --build C:\MITK_build\MITK-build --config Release --target MitkVesselFM
    • cmake --build C:\MITK_build\MITK-build --config Release --target MitkVesselFMUI
  3. Run the plugin in MITK

    1. Load a file you want to segment
    2. Create a new segmentation label
    3. In the segmentation window choose 3D
    4. For the proposed nnInteractive_>vesselFM workflow: Choose nnInteractive
    5. Intialize and install dependencies(first time use only)
    6. Use the prompt tool to segment the aneurysm sac (i.e. via point prompt)
    7. Confirm segmentation
    8. Choose the vesselFM plugin
    9. Restrict mask around roi
    10. Choose threshold or leave at default as well as pixel number for the padding box around the roi from nnInteractive
    11. Preview segmentation
    12. Confirm segmentation
    13. Get volume metrics

About

This repository integrates the essential code files needed for setting up vesselFM as a working plugin in GitHub.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • C++ 93.0%
  • CMake 6.5%
  • C 0.5%