NA-MIC Project Weeks

Update the Chest Imaging Platform extenstion to support Slicer 5

Key Investigators

Rudolf Bumm (KSGR)
Raul San Jose Estepar (Brigham)
Andras Lasso (PerkLab)
Steve Pieper (Kitware)

Project Description

The Chest Imaging Platform (CIP) is an extension to 3D Slicer.

It integrates:

chest image processing functionality as a toolkit by exposing the CLIs
Slicer specific modules to provide user-friendly chest CT quantitative solutions
Visualization of scale-space particles and labelmaps
Integrated workflows to end-to-end clinical evaluation

In the current preview versions of 3D Slicer (4.13.0) parts of CIP fail to load the following CIP modules because Slicer’s “Editor” module has been removed.

CIP_CalciumScoring

CIP_RVLVRatio

CIP_LesionModel

CIP_Calibration

CIP_MIPViewer

CIP_BodyComposition

CIP_ParenchymaSubtypeTrainingLabelling

CIP_ParenchymaAnalysis

CIP_PAARatio

CIP_AVRatio

CIP_InteractiveLobeSegmentation

The CIP extension currently uses legacy editor module, and charts infrastructure (instead of Segment editor, Plots, and Tables modules).

Objective

Replace the usage of the “Editor” module in CIP by something different, preferably the SegmentEditor.
Replace labelmaps with segmentations
Replace outdated fiducial calls (exceptions)
Replace the charts infrastructure

Approach and Plan

A fork and branch of SlicerCIP was created at https://github.com/rbumm/SlicerCIP/tree/Branch_CIPCompatSlicer5 to be used as the source base, all changes will later be included in a PR to https://github.com/acil-bwh/SlicerCIP.

Resolve compatibility problems step by step

Deactivate non-working or outdated modules in Slicer 5

Considerations

Progress and Next Steps

01/06/22:

Removing the “Editor” related imports from “Scripted/CIP_/CIP/ui/init.py” results in a complete CIP-startup in Slicer 4.13.0 without initial error messages.

A github search revealed that “Editor” calls are being made from three of the above modules:

CIP_Calibration
CIP_ParenchymaSubtypeTrainingLabelling
CIP_BodyComposition

01/11/22:

CIP_Calibration is probably outdated.
CIP_ParenchymaSubtypeTrainingLabelling is outdated, probably redundant.
CIP_BodyComposition is needed, but probably much better realized with AI segmentation

01/17/22: 12pm Meeting with Raul, Andras, Steve and Rudolf (Discord Red Slice)

Incompabilities between 4.13 and CIP seem to be caused by CIPLibrary
CIP, it’s CLI functions and their history were demonstrated by @Raul
CIP_Calibration is not outdated.
CIP_ParenchymaSubtypeTrainingLabelling should be kept.
CIP_BodyComposition is needed and should be kept
CIP Toolkit functions should be included in a future SlicerCIP release
A CIPLibrary compatibility branch “4.13” or “5.0” will be created
CIP GitHub write access was requested for @Andras to support merging, @Raul agreed
As labelmaps are used throughout CIP in nearly every module we will discuss a Slicer-based solution for that problem tomorrow

01/18/22: 12pm Meeting with Raul, Andras, Rudolf (Discord Red Slice)

good follow up meeting
@rbumm demonstrated the build process
a first module (Calibration) has been adapted to 4.13 and now works with the Segment Editor instead of Editor
ParenchymaSubtypeTrainingLabelling was demonstrated by Raul
we decided to prioritize certain modules during CIP adaptation
ParenchymaAnalysis probably the next to go
Andras promised to look into CLI modules and letting them use segmentations instead of labelmaps as an input maybe automatically
next meeting will be Thursday 11 am Red Slice

Final presentation start here

01/20/22: 11am-12:30pm Meeting with Raul, Andras, Rudolf (Discord Red Slice)

The Chest Imaging Platform extension’s “Calibrate” module has been finalized for Slicer 5 together with @lassoan and was demonstrated by @rbumm
see this SlicerCIP GitHub fork and this commit
will serve as a skeleton to adopt the other CIP modules and is now fully functional
The “Calibration” widget uses a newly created segmentation for each volume instead of labelmaps
the segmentation can be edited in the embedded Segment Editor
before pressing “Calibrate” the segmentation is converted to a labelmap for the actual calibration in the logic()
in addition, we developed a strategy on how to transform the other CIP modules to Slicer 5
non-working modules will be excluded from CMake to provide only those functional to the community
@Raul demonstrated new vessel segmentation techniques and answered many questions
@Raul agreed on giving @lassoan write access to the SlicerCIP GitHub repository

Illustrations

New Slicer 5 CIP “Calibration” module with embedded “SegmentEditor” instead of the old “Editor”:

Background and References

https://chestimagingplatform.org/

https://discourse.slicer.org/t/exporting-csv-with-parenchyma-analysis-module/10697/58?u=rbumm