Ground glass nodule benign and malignant classification method based on two-way three-dimensional convolutional neural network

Abstract

The invention discloses a ground glass nodule benign and malignant classification method based on a two-way three-dimensional convolutional neural network. The method comprises the steps of 1, collecting participants; 2, acquiring an image; 3, preprocessing the image; and 4, training a deep learning model. According to the ground glass nodule benign and malignant classification method based on the two-way three-dimensional convolutional neural network, data obtained from a PET image and a CT image is developed and fused through deep learning to serve as the two-way 3D-CNN of an input stream so as to be used for classifying benign lesions and malignant lesions in the GGN. According to the network, an end-to-end working process can be realized in a classification process without professional image analysis software and too much human intervention; in addition, the classification accuracy of the network is higher than that of the qualified nuclear medicine practitioners; and then the CNN may contribute to benign and malignant classification and clinical management of the GGN in the absence of well-trained and experienced film readers.

Description

technical field [0001] The invention belongs to the technical field of image analysis of pulmonary ground-glass nodules, and in particular relates to a method for classifying benign and malignant ground-glass nodules based on a two-way three-dimensional convolutional neural network. Background technique [0002] With the popularization of low-dose CT and the screening of the new crown epidemic, the incidence of early lung adenocarcinoma manifested as ground-glass opacity nodules (GGN) has increased rapidly. In the management of patients with GGN, estimating the clinically predictive probability of malignancy should be an important early step. The guideline recommends that for temporarily uncertain GGNs, the dynamic changes of GGNs can be classified by CT follow-up. However, some benign GGN and early lung adenocarcinoma can remain stable for a long period of time, making it difficult to differentiate. Moreover, long-term follow-up often brings panic and anxiety to patients....

AI Technical Summary

Problems solved by technology

But to the best of our knowledge, the use of a three-dimensional convolutional neural network (3D-CNN) to analyze PET / CT images, and a two-way CNN (data obtained from PET and CT as separate input streams) for benign and malignant classification of GGN have not been developed.

Images