Deep learning segmentation system for hepatic vein and hepatic portal vein

Abstract

The invention provides a deep learning segmentation system for hepatic veins and hepatic portal veins. The deep learning segmentation system comprises a coding module, a sequence attention association fusion module, an inter-slice-graph association module and a decoding module. An inter-slice-graph association module designed based on a graph neural network captures richer and higher-order relationships between adjacent CT sequences and improves the continuity of blood vessel segmentation; a sequence attention association fusion module designed based on an attention mechanism associates and fuses information of different dimensions, and the fine blood vessel segmentation accuracy is improved; in addition, the invention further provides a loss function combined with a blood vessel edge measurement constraint term, and the blood vessel edge is optimized and adjusted.

Description

technical field [0001] The present invention relates to the field of medical image processing, and more specifically, to a deep learning segmentation system for hepatic veins and hepatic portal veins. Background technique [0002] The incidence of liver disease is gradually increasing, which is largely harmful to human health. It is currently one of the diseases with the highest mortality in China. Many liver diseases need to be treated by liver resection. Before the operation, doctors need to segment the liver according to the trend of the hepatic vein and hepatic portal vein, so as to remove the diseased area of ​​the liver as accurately as possible and preserve the normal area. In clinical research, most medical imaging diagnoses are completed by doctors' naked eye observation and clinical experience. However, it is time-consuming, labor-intensive and subjective to conduct countless medical image analysis, which is likely to cause missed and misdiagnosed phenomena. There...

AI Technical Summary

Problems solved by technology

However, cutting blocks will limit the maximum receptive field that the network can achieve, resulting in the loss of certain global information. The target of the hepatic vein and hepatic portal vein itself is much larger than the cut block. In this case, it is difficult for the network to learn the overall structural information of blood vessels. , has a huge impact on the overall segmentation of the hepatic vein and hepatic portal vein

Images