Liver CT image multi-lesion classification method based on sample generation and transfer learning

Abstract

The invention discloses a liver CT image multi-lesion classification method based on sample generation and transfer learning. The method mainly solves the problem that an existing method is not high in liver multi-lesion detection performance. The implementation scheme is as follows: dividing a data set; respectively constructing a liver organ segmentation network and a liver lesion detection network; based on the deep convolution generative adversarial network, constructing a liver cyst sample generation network and a liver hemangioma sample generation network, and respectively generating newliver cyst and liver hemangioma samples; constructing a liver lesion classification network; subjecting a liver CT image to be detected firstly to organ segmentation by using a liver organ segmentation network, then subjecting a segmentation result to lesion detection by using a liver lesion detection network, and finally classifying detected lesions by using a liver lesion classification network. According to the invention, imbalance of different types of sample sizes is relieved, the lesion classification performance is improved, and the method can be used for positioning and qualifying various lesions such as liver cancer, liver cyst and hepatic hemangioma existing in the liver CT image.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a method for classifying liver lesions, which can be used to locate and characterize various lesions such as liver cancer, hepatic cysts, and hepatic hemangiomas in liver CT images. Background technique [0002] As the largest digestive gland in the human body, the liver plays an irreplaceable role. However, there are many kinds of diseases. For example, liver cancer is one of the common malignant tumors. The incidence rate is increasing year by year. The early manifestations are usually not obvious. Epigastric discomfort, pain, fever, and fatigue will only appear when the tumor is large or the course of the disease progresses to the middle and late stages. , nausea, loss of appetite and other symptoms. Imaging diagnosis is of great significance to the diagnosis and curative effect evaluation of liver cancer, liver cyst, hepatic hemangioma and other diseases. In ...

AI Technical Summary

Problems solved by technology

[0005] Among the current research methods for lesion detection and classification, the traditional image processing method through artificial design and lesion-related features, and the deep learning method through convolutional neural network learning and automatic feature extraction are mostly two methods, but these methods are large. Most research directions focus on how to effectively detect a specific lesion, and cannot accurately detect and classify multiple types of lesions at the same time

Images