Medical OOD Segmentation Detection

Medical Segmentation Detection

Type: Master CS

Period: TBD

Student: (Unassigned)

If you are interested please contact :

dr. A. Briasouli PhD (Alexia) Assistant Professor

+31534896985

 a.briassouli@utwente.nl

Building: Zilverling 4049

Personal page

The 2024 benchmark MedSegBench: A comprehensive benchmark for medical image segmentation in diverse data modalities has been tested successfully with SoA architectures such as UNet with ResNet. However, novel medical segmentation Foundation Models (FMs) are now available, which may provide rich feature representations.

• Medical Segmentation FMs should be tested on MedSegBench for segmentation performance and be examined for generalizability in realistic OOD scenarios.
• OOD detection methods should be applied to detect OOD samples.
  • Pytorch OOD
  • OODEEL
  • Post-hoc out-of-distribution detection for cardiac MRI segmentation (2025)
  • Optional: Retraining without OOD samples (fine tuning) may take place for performance comparison.

 REFERENCES:

1.  MedSegBench: A comprehensive benchmark for medical image segmentation in diverse data modalities
2. OpenMIBOOD: Open Medical Imaging Benchmarks for Out-Of-Distribution Detection
3. MedSAM
4. MedSAM2
5. MedSAM3
6. VISION-MAE