[B] OOD detection on MediCLIP

Master Assignment

[B] OOD detection on MediCLIP

Type: Master EE/CS/ITC

Period: TBD

Student: (Unassigned)

If you are interested please contact :

Background:

MediCLIP,adapts the CLIP model to few-shot medical image anomaly detection through self-supervised fine-tuning. MediCLIP designs a series of medical image anomaly synthesis tasks to simulate common disease patterns in medical imaging, transferring the powerful generalization capabilities of CLIP to the task of medical image anomaly detection. When

only few-shot normal medical images are provided, MediCLIP achieves state-of-the-art performance in anomaly detection and location compared to other methods.

MediCLIP uses its own anomaly detection method, however a large number of OOD detection methods exist, encoded in PyTorchOOD. This thesis should compare OOD detection methods on the MediCLIP dataset and other medical image OOD datasets. XAI can be used to interpret or improve results using Explanation Guided Learning.

Resources:

  1. MediCLIP: Adapting CLIP for Few-shot Medical Image Anomaly Detection https://papers.miccai.org/miccai-2024/504-Paper0333.html 
  2. PyTorch-OOD: A Library for Out-of-Distribution Detection Based on PyTorch https://openaccess.thecvf.com/content/CVPR2022W/HCIS/html/Kirchheim_PyTorch-OOD_A_Library_for_Out-of-Distribution_Detection_Based_on_PyTorch_CVPRW_2022_paper.html 
  3. Investigation of out-of-distribution detection across various models and training methodologies https://www.sciencedirect.com/science/article/pii/S0893608024002120 
  4. MOOD 2020: A Public Benchmark for Out-of-Distribution Detection and Localization on Medical Images https://pubmed.ncbi.nlm.nih.gov/35468060/ 
  5. A Benchmark of Medical Out of Distribution Detection https://openreview.net/forum?id=oUg5rC_95OM