Liver CT tumor classification system and classification method

Abstract

The invention discloses a liver CT tumor classification system and classification method. The liver CT tumor classification system comprises a suspicious liver tumor detection module, a candidate liver tumor segmentation module and a candidate liver tumor classification module. The suspicious liver tumor detection module adopts a liver spectrum and manual correction combined method to segment out liver parenchyma, then a variable annular filter is adopted to carry out suspicious liver tumor detection to obtain candidate liver tumors, and the center of the variable annular filter is determined through gray scale weight distance conversion to serve as a candidate liver tumor seed point; the candidate liver tumor segmentation module completes candidate liver tumor segmentation by using a 3D region growing algorithm; the candidate liver tumor classification module is used for carrying out rough false positive reduction on candidate liver tumors by adopting an empirical threshold method, and dividing the candidate liver tumors into tumors and other liver tissues by taking an SVM algorithm as a classifier. The liver CT tumor detection sensitivity is high, and the classification accuracy is effectively improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a liver CT tumor classification system and a classification method. Background technique [0002] As the largest glandular organ in the human body, the liver plays an important role in human metabolism. The liver is located in the abdomen of the human body, adjacent to many important organs, and has a sufficient blood supply. Once the disease is triggered, it is very easy to transfer to the surrounding organs, and it is also easily affected by other organs. In addition, as the second leading cause of death, liver cancer has an annual incidence rate of 26.39 per 100,000 people in China. The lack of medical classification is an important reason for the high mortality rate. Compared with MR, X-ray and other radiation techniques, CT images have the advantages of fast imaging, high contrast, and low cost. Therefore, they are widely used in the detection and classification of...

AI Technical Summary

Problems solved by technology

[0003] Since the liver is adjacent to multiple organs, and the gray level of the liver image varies from person to person and from device to device, at the same time, damage close to the liver boundary will cause the liver boundary to be blurred, resulting in very complicated extraction of the liver region. Accurate detection from the liver region Suspicious targets are very difficult. Suspicious areas here include two key tissues: blood vessels and cancerous tissues. In the process of extracting suspicious areas, shape features, image density, and texture features have been widely used. However, liver tumors in different periods Density and texture information will have different performances. Therefore, it is difficult to accurately extract suspicious objects purely relying on basic threshold information; and methods based on shape information, because the position of tumors in the liver will change with time, it is very difficult It is difficult to achieve accurate detection of tumors; the last part is the classification of candidate tumors or diseases. At this stage, the traditional support vector machine SVM classification algorithm is usually used to classify candidate areas into liver tumor tissue and other liver tissues. Traditional SVM classification algorithms are mostly based on The Gaussian radial basis kernel is a kernel function that tends to be local, and its ability to expand outward will decrease with the increase of the amplitude parameter σ. It is difficult to train and classify large-scale samples, so the accuracy of liver tumor classification cannot be guaranteed. sex

Images