NA-MIC Project Weeks

Back to Projects List

Integrate MRSegmentator and DeepSpA models into mhub.ai

Key Investigators

• Felix Dorfner (Charité Universitätsmedizin Berlin, Germany)
• Hartmut Häntze (Charité Universitätsmedizin Berlin, Germany)
• Keno Bressem (Technical University of Munich, Germany)

Presenter location: In-person

Project Description

This project will aim to integrate two models into mhub.ai.

1. MRSegmentator:

• Is a segmentation model that can accurately segment 40 organs and structures in human MRI scans of the abdominal, pelvic and thorax regions. The model works on different sequence types, including T1- and T2-weighted, Dixon sequences and even CT images.
• Paper: https://arxiv.org/abs/2405.06463

2. DeepSpA:

• Is classification model that incorporates anatomical awareness to detect radiographic sacroiliitis. Detecting radiographic sacroiliitis plays an essential role in diagnosing and classifying axial Spondyloarthritis (axSpA).
• Paper: https://arxiv.org/abs/2405.07369

Objective

1. Objective: Working implementation of MRSegmentator into the mhub.ai platform.
2. Objective: Working implementation of DeepSpA into the mhub.ai platform

Approach and Plan

Progress and Next Steps

During Project Week

• MRSegmentator is wrapped in the MhubAI framework and ready to be tested
  • Segmentations are registered as DICOM-SEG and can easily compared to other segmentation models
• DeepSpA is wrapped in the MhubAI framework and ready to be tested
  • Model produces both visual and classification outputs, both are organized and saved by mhub

After Project Week

• Complete the testing process and publish both models on MHub.ai

Illustrations

Illustrations of both models can be seen on their respective GitHub pages, linked below:

Background and References

MRSegmentator:

• Code: https://github.com/hhaentze/MRSegmentator
• Paper: https://arxiv.org/abs/2405.06463

DeepSpA:

• Code: https://github.com/FJDorfner/Anatomy-Aware-Classification-axSpA
• Paper: https://arxiv.org/abs/2405.07369