Deep learning-based pulmonary fissure segmentation and integrity assessment method and system

Abstract

The invention discloses a deep learning-based pulmonary fissure segmentation and integrity assessment method and system. Compared with a traditional method, the method has the advantages that the pulmonary fissure segmentation precision and calculation efficiency are obviously improved, and full-automatic pulmonary fissure segmentation and pulmonary fissure integrity assessment are realized. The main steps of the pulmonary fissure segmentation and integrity assessment method comprise: constructing a pulmonary fissure segmentation data set; training a pulmonary fissure segmentation model basedon the full convolutional neural network; predicting a pulmonary fissure area and identifying to obtain left lung oblique fissure, right lung oblique fissure and right lung horizontal fissure; estimating complete pulmonary fissure; evaluating the degree of pulmonary fissure integrity. A full convolutional neural network is adopted; end-to-end pulmonary fissure model training and prediction are realized; the method has the advantages that the prediction speed is high without manual intervention, a segmentation framework from coarse to fine is adopted, the problem that the number of category labels is extremely unbalanced when a full convolutional neural network is used for performing a segmentation task is solved, false positive generated by pulmonary fissure segmentation is removed by introducing pulmonary fissure segmentation, and the pulmonary fissure integrity assessment is more accurate.

Description

technical field [0001] The present invention relates to the field of medical image processing, in particular to a method and system for lung fissure segmentation and integrity assessment based on deep learning. Background technique [0002] Lung fissure plays a very important role in the detection, classification and evaluation of lung diseases. Therefore, it is of great significance to accurately locate the pulmonary fissure area and segment the lung fissure in the diagnosis of lung diseases. In CT images, the pulmonary fissure appears as a curve with little curvature change in the two-dimensional slice structure, but appears as a ribbon structure or planar structure in the three-dimensional structure. In clinical diagnosis, understanding the structural characteristics of lung fissures is helpful for the localization of lung lesions and the quantitative assessment of lung diseases. However, it is very difficult to complete automatic pulmonary fissure segmentation in CT ima...

AI Technical Summary

Problems solved by technology

[0003] A large number of detection and segmentation methods (such as the segmentation of airways, blood vessels and lungs) are related to lung CT images and have developed maturely, but the detection and segmentation methods of lung fissures are still under study. Currently, lung fissure segmentation generally focuses on computational geometry. The method of automatic detection and segmentation of lung fissures, but this scheme has limitations, this method can achieve good results only under "favorable" conditions, that is, its detection requires preconditions; the execution efficiency of this scheme is low

Images