Aiming at the following problems encountered by Multiple Supervised Fully Convolutional Networks (MSFCN ) in liver tumor segmentation: rich blood vessels around the liver, complex organs, and inconspicuous boundaries, which have a great impact on accurate tumor segmentation, plus The tumor boundary is not clear and the segmentation is difficult. In this paper, we propose a method to automatically segment liver and tumor in CT abdominal images using Cascaded Fully Convolutional Networks (CFCN) and 3D Conditional Random Fields (CRF). We train and cascade two fully convolutional neural networks for combined segmentation of the liver and its tumors. In the first step, our trained FCN segmented the liver from abdominal CT images as the input of the second FCN, which removed the influence of the environment around the liver. The second FCN segmented the tumor region only based on the liver obtained in the first step. Finally, 3D conditional random field is used to optimize the segmented image obtained by cascaded FCN, and the edge information of liver tumor image is extracted to solve the problem of unclear boundary of liver tumor and complete the segmentation.
For the convolutional neural network in the process of image segmentation, there are problems such as loss of detailed information and global features of the image, and the segmentation results of the full convolutional neural network in the image segmentation process are not refined and lack spatial consistency. This paper proposes a liver tumor segmentation method based on Multiple Supervised Fully Convolutional Networks (MSFCN). In the framework of a fully convolutional network, a supervised side output layer is added to the convolutional layer to guide multi-scale feature learning, which can better capture the local and global features of the image. Experiments are carried out on the LiTS competition dataset, and the experimental results show that the proposed method can significantly improve the segmentation accuracy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.