Pancreas segmentation with probabilistic map guided bi-directional recurrent UNet

  • 2022-08-11 16:33:22
  • Jun Li, Xiaozhu Lin, Hui Che, Hao Li, Xiaohua Qian
  • 0

Abstract

Pancreas segmentation in medical imaging data is of great significance forclinical pancreas diagnostics and treatment. However, the large populationvariations in the pancreas shape and volume cause enormous segmentationdifficulties, even for state-of-the-art algorithms utilizingfully-convolutional neural networks (FCNs). Specifically, pancreas segmentationsuffers from the loss of spatial information in 2D methods, and the highcomputational cost of 3D methods. To alleviate these problems, we propose aprobabilistic-map-guided bi-directional recurrent UNet (PBR-UNet) architecture,which fuses intra-slice information and inter-slice probabilistic maps into alocal 3D hybrid regularization scheme, which is followed by bi-directionalrecurrent network optimization. The PBR-UNet method consists of an initialestimation module for efficiently extracting pixel-level probabilistic maps anda primary segmentation module for propagating hybrid information through a 2.5DU-Net architecture. Specifically, local 3D information is inferred by combiningan input image with the probabilistic maps of the adjacent slices intomultichannel hybrid data, and then hierarchically aggregating the hybridinformation of the entire segmentation network. Besides, a bi-directionalrecurrent optimization mechanism is developed to update the hybrid informationin both the forward and the backward directions. This allows the proposednetwork to make full and optimal use of the local context information.Quantitative and qualitative evaluation was performed on the NIH Pancreas-CTdataset, and our proposed PBR-UNet method achieved better segmentation resultswith less computational cost compared to other state-of-the-art methods.

 

Quick Read (beta)

loading the full paper ...