Pulmonary nodule CT image classification method based on adaptive selection double-source domain heterogeneous transfer learning

Abstract

The invention belongs to the technical field of CT image classification, and particularly relates to a pulmonary nodule CT image classification method based on adaptive selection double-source domain heterogeneous transfer learning. According to the method, preoperative auxiliary diagnosis of PT and LA is carried out on an SPSN patient by using a CT image. The method comprises two parts: (1) feature extraction of double-source domain heterogeneous transfer learning based on adaptive selection, and (2) construction of an integrated classifier based on a sparse Bayesian extreme learning machine. The self-adaptive selection-based dual-source domain heterogeneous transfer learning model adaptively determines the matching weight of each pair of feature maps between a source network and a target network and the matching weight of each pair of convolution blocks between a source network feature block and the target network by designing a self-adaptive selection-based dual-source domain feature matching network; according to the method, features beneficial to target task learning and feature migration destinations in a source network are automatically selected, so that training of a target network is restrained, and the robustness of the target network under the condition of small samples is improved.

Description

technical field [0001] The invention belongs to the technical field of CT image classification, and in particular relates to a lung nodule CT image classification method based on adaptive selection of dual-source domain heterogeneous transfer learning. Background technique [0002] With the development of Computed Tomography (CT) technology, the detection rate of Solitary Pulmonary Solid Nodule (SPSN) has been greatly improved. SPSN is the English abbreviation of Solitary Solid Pulmonary Nodule, PT is the English abbreviation of Pulmonary Tuberculosis; LA is the English abbreviation of Lung Adenocarcinoma, and Pulmonary Tuberculosis (PT) is a typical histopathological manifestation of benign SPSN; , Lung adenocarcinoma (Lung Adenocarcinomas, LA) is the most common histological type of lung cancer. In clinical practice, LA patients should adopt more aggressive treatment regimens to improve prognosis; while PT patients should avoid unnecessary treatment procedures (such as su...

AI Technical Summary

Problems solved by technology

[0005] Although the above method makes the model have a certain effect in the case of heterogeneous data sources, there are two problems: ①In a source network, different features have different importance to the target task, and some features are even redundant. the rest

Negative transfer may occur when redundant features are transferred to the target network

② Determining how the features in the source network migrate to the target network based on experience alone will consume a lot of manpower and material resources, and the result may not be optimal

Images