eduzhai > Applied Sciences > Engineering >

Renal Cell Carcinoma Detection and Subtyping with Minimal Point-Based Annotation in Whole-Slide Images

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Obtaining a large amount of labeled data in medical imaging is laborious andtime-consuming, especially for histopathology. However, it is much easier andcheaper to get unlabeled data from whole-slide images (WSIs). Semi-supervisedlearning (SSL) is an effective way to utilize unlabeled data and alleviate theneed for labeled data. For this reason, we proposed a framework that employs anSSL method to accurately detect cancerous regions with a novel annotationmethod called Minimal Point-Based annotation, and then utilize the predictedresults with an innovative hybrid loss to train a classification model forsubtyping. The annotator only needs to mark a few points and label them arecancer or not in each WSI. Experiments on three significant subtypes of renalcell carcinoma (RCC) proved that the performance of the classifier trained withthe Min-Point annotated dataset is comparable to a classifier trained with thesegmentation annotated dataset for cancer region detection. And the subtypingmodel outperforms a model trained with only diagnostic labels by 12 in termsof f1-score for testing WSIs.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×