eduzhai > Applied Sciences > Engineering >

Multi-level colonoscopy malignant tissue detection with adversarial CAC-UNet

  • Save

... pages left unread,continue reading

Document pages: 15 pages

Abstract: The automatic and objective medical diagnostic model can be valuable toachieve early cancer detection, and thus reducing the mortality rate. In thispaper, we propose a highly efficient multi-level malignant tissue detectionthrough the designed adversarial CAC-UNet. A patch-level model with apre-prediction strategy and a malignancy area guided label smoothing is adoptedto remove the negative WSIs, with which to lower the risk of false positivedetection. For the selected key patches by multi-model ensemble, an adversarialcontext-aware and appearance consistency UNet (CAC-UNet) is designed to achieverobust segmentation. In CAC-UNet, mirror designed discriminators are able toseamlessly fuse the whole feature maps of the skillfully designed powerfulbackbone network without any information loss. Besides, a mask prior is furtheradded to guide the accurate segmentation mask prediction through an extramask-domain discriminator. The proposed scheme achieves the best results inMICCAI DigestPath2019 challenge on colonoscopy tissue segmentation andclassification task. The full implementation details and the trained models areavailable at this https URL.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×