eduzhai > Applied Sciences > Engineering >

ARPM-net A novel CNN-based adversarial method with Markov Random Field enhancement for prostate and organs at risk segmentation in pelvic CT images

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 23 pages

Abstract: Purpose: The research is to develop a novel CNN-based adversarial deeplearning method to improve and expedite the multi-organ semantic segmentationof CT images, and to generate accurate contours on pelvic CT images. Methods:Planning CT and structure datasets for 120 patients with intact prostate cancerwere retrospectively selected and divided for 10-fold cross-validation. Theproposed adversarial multi-residual multi-scale pooling Markov Random Field(MRF) enhanced network (ARPM-net) implements an adversarial training scheme. Asegmentation network and a discriminator network were trained jointly, and onlythe segmentation network was used for prediction. The segmentation networkintegrates a newly designed MRF block into a variation of multi-residual U-net.The discriminator takes the product of the original CT and theprediction ground-truth as input and classifies the input into fake real. Thesegmentation network and discriminator network can be trained jointly as awhole, or the discriminator can be used for fine-tuning after the segmentationnetwork is coarsely trained. Multi-scale pooling layers were introduced topreserve spatial resolution during pooling using less memory compared to atrousconvolution layers. An adaptive loss function was proposed to enhance thetraining on small or low contrast organs. The accuracy of modeled contours wasmeasured with the Dice similarity coefficient (DSC), Average Hausdorff Distance(AHD), Average Surface Hausdorff Distance (ASHD), and relative VolumeDifference (VD) using clinical contours as references to the ground-truth. Theproposed ARPM-net method was compared to several stateof-the-art deep learningmethods.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...