A method and system for automatic scoring of semantic features based on CT images of pulmonary nodules

Abstract

The invention discloses a method and system for automatically scoring semantic features based on pulmonary nodule CT images. The method includes: extracting computer features of pulmonary nodule CT images through different algorithms to generate heterogeneous features corresponding to different semantic features; Heterogeneous features use a multi-task regression algorithm including block constraints and pixel sparse constraints to obtain shared features between semantic features and independent features of each semantic feature; according to shared features between semantic features and independent features of each semantic feature The combined mapping relationship automatically scores each semantic feature. The present invention can effectively extract the effective features of each semantic feature, and obtain the memory relationship between each semantic feature, thereby scoring the semantic features, realizing in-depth analysis of pulmonary nodule information, and supporting diagnostic reports with complex content and medical images. Retrieval provides effective assistance for pulmonary nodule CT image analysis.

Description

technical field [0001] The invention relates to the technical field of medical images, in particular to an automatic scoring method and system for semantic features based on CT images of pulmonary nodules. Background technique [0002] Semantic features of pulmonary nodules, such as spiculation and lobulation, are generally used to describe the phenotype of pulmonary nodules in radiology reports. Studies have shown that in the different diagnosis of pulmonary nodules CT images, high-level texture features, such as spiculation and nodular parenchyma, are important diagnostic basis for identifying benign and malignant nodules. Past studies have proposed several classification schemes for computer-predicted semantic features. For example, the automatic classification of 51 nodules with glitch sign and 204 nodules without glitch sign can be realized through the descriptor; or the parenchymal grade of pulmonary nodules can be classified by using low-level contrast features. Pre...

AI Technical Summary

Problems solved by technology

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method and system for automatic scoring of semantic features based on CT images of pulmonary nodules, aiming at solving the problem of being unable to associate various semantic features of pulmonary nodules in the prior art. The existing analysis methods are simple, and it is impossible to carry out quantitative and automatic analysis of high-level semantic features.

Images