Full-automatic liver tumor segmentation method based on a two-way three-dimensional convolutional neural network

Abstract

The invention provides a liver tumor image segmentation method based on a two-way three-dimensional convolutional neural network. The method comprises the following steps: preparing a data set; performing filtering and standardized preprocessing operation on the original CT image in the data set; training a two-way three-dimensional convolutional neural network with a parallel path structure; segmenting a liver tumor in the CT image by using the trained two-way three-dimensional convolutional neural network, and generating a probability graph of a tumor segmentation result; based on the two-way three-dimensional convolutional neural network, tumor segmentation of the liver CT image is fully automatically realized, the accuracy is high, and the speed is high.

Description

technical field [0001] The present invention relates to the technical field of computer-aided diagnosis, in particular to a fully automatic liver tumor segmentation method based on a two-way three-dimensional convolutional neural network. Background technique [0002] The liver is the largest internal organ of the human body, the most vigorous metabolism of the human body, and the most important detoxification organ of the human body. At the same time, the liver is also one of the organs with a high incidence of human tumors, and the incidence of malignant tumors in the liver is much higher than that of benign tumors. According to statistics, worldwide, the incidence of liver cancer is increasing at a rate of about 700,000 people per year, and the incidence and mortality of liver cancer in China account for more than 50% of the world. The latest National Cancer Report 2018 released by the National Cancer Center pointed out that among cancer patients in my country, the incid...

AI Technical Summary

Problems solved by technology

However, obtaining tumor regions by manual segmentation on hundreds of CT slices is very tedious and time-consuming, and introduces human subjectivity, and the knowledge and experience level of experts are important factors affecting the segmentation accuracy

In the research on the automatic segmentation algorithm of liver tumors, there are mainly the following difficulties: First, the edges between the liver and its adjacent organs, as well as some muscle, fat and other human tissues in the abdominal cavity are very compact, and the contact area between them is relatively small. Large, but their CT values ​​are not much different, so the segmentation of the liver edge brings a lot of interference; secondly, the low contrast between the tumor and the healthy tissue inside the liver leads to small observable changes, making its boundaries blurred ; Furthermore, the imaging manifestations of liver tumors are complex and variable in size, shape, location, etc., and there are large differences between the same patient and between patients

Images