Integrated ResNet-NRC method for dividing sample space based on lung tumor image

Abstract

The invention discloses an integrated ResNet-NRC method for dividing a sample space based on a lung tumor image, and the method comprises the following steps: collecting medical image information of the same lung tumor in three dimensions, i.e., a three-mode data set CT, a PET and a PET / CT; dividing sample feature spaces of three modes according to the three-mode data set CT, PET and PET / CT; constructing residual neural network models of the three modes according to the sample feature spaces of the three modes, namely a base classifier; and combining the three base classifiers by adopting a relative majority voting method to form a final classification identification result. According to the method, the classification accuracy is excellent, the conditions of high accuracy and large difference of a base classifier are met, the optimization problem of high-dimensional data can be effectively solved, the specificity, sensitivity and other evaluation indexes are high, and the method has good robustness and generalization ability.

Description

technical field [0001] The present invention relates to the technical field of medical image recognition, and more specifically relates to an integrated ResNet-NRC method based on lung tumor image division sample space. Background technique [0002] Medical imaging methods are widely used in the diagnosis of lung tumors, including X-ray imaging, computerized tomography (CT), positron emission computed tomography (PET) and magnetic resonance imaging (Magnetic Resonance Imaging, MRI), etc., PET is mainly based on tracers to selectively reflect the metabolism of tissues and organs. It reflects the physiological, pathological, biochemical and metabolic changes of human tissues at the molecular level, and is suitable for the study of human physiological functions. 18F-FDG PET / CT combines the advantages of PET and CT, and integrates anatomical images (CT) and functional metabolic images (PET) on the same machine, which has both fine anatomical structures and rich physiological, b...

AI Technical Summary

Problems solved by technology

[0004] However, the existing ensemble learning methods cannot meet the conditions of high accuracy and large differences for the base classifier, and their specificity and sensitivity for medical image recognition cannot meet the standards.

Images