SemicNet: a semicircular network for the segmentation of the liver and its lesions
by Zhihua Zheng; Victor S. Sheng; Lei Wang; Zhi Li; Xuefeng Xi; Zhiming Cui
International Journal of Sensor Networks (IJSNET), Vol. 35, No. 3, 2021

Abstract: The traditional neural network used for medical image segmentation was not clear on the network depth, the importance of different depths and the rationality of jump connection. In view of these problems, we propose a convenient and efficient liver and lesion segmentation system, which uses a double-layer codec semi-circular network to combine the deep and shallow semantic information through dense jump connection, which is easier to be processed by the optimiser; The transition zone between liver and lesion segmentation is designed so that the result of liver segmentation can be effectively transmitted to lesion segmentation; We believe that the selection of complementary loss function combination for in-depth supervision can effectively receive the anti-propagation gradient signal and obtain more regularisation effects. Finally, in terms of liver segmentation, in addition to the model with lower accuracy than multiple loss functions for joint decision-making, all other evaluation indexes, including lesions, exceeded the fusion results of multiple models.

Online publication date: Wed, 31-Mar-2021

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Sensor Networks (IJSNET):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com