eduzhai > Applied Sciences > Engineering >

Universal Model for Multi-Domain Medical Image Retrieval

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: Medical Image Retrieval (MIR) helps doctors quickly find similar patients data, which can considerably aid the diagnosis process. MIR is becomingincreasingly helpful due to the wide use of digital imaging modalities and thegrowth of the medical image repositories. However, the popularity of variousdigital imaging modalities in hospitals also poses several challenges to MIR.Usually, one image retrieval model is only trained to handle images from onemodality or one source. When there are needs to retrieve medical images fromseveral sources or domains, multiple retrieval models need to be maintained,which is cost ineffective. In this paper, we study an important but unexploredtask: how to train one MIR model that is applicable to medical images frommultiple domains? Simply fusing the training data from multiple domains cannotsolve this problem because some domains become over-fit sooner when trainedtogether using existing methods. Therefore, we propose to distill the knowledgein multiple specialist MIR models into a single multi-domain MIR model viauniversal embedding to solve this problem. Using skin disease, x-ray, andretina image datasets, we validate that our proposed universal model caneffectively accomplish multi-domain MIR.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×