Vertebral body bone mineral density classification method based on image omics and deep learning feature fusion

Abstract

The invention relates to the technical field of medical image segmentation and image classification, in particular to a centrum bone mineral density classification method based on radiomics and deep learning feature fusion, centrum under a CT image is divided into an osteoporosis group, a low bone mass group and a normal group, and the centrum bone mineral density classification method comprises the following steps: S1, establishing a centrum segmentation network based on CRF and attention guidance, and obtaining centrum cancellous bone masks L1 and L2; s2, deep learning feature extraction is performed on the feature map fused by the L1 and the L2 through GCAM-Net, and radiomics feature extraction is performed by using CT images and masks of the L1 and the L2; and S3, extracting an optimal feature set from the deep learning features by using a differential evolution algorithm, carrying out feature screening on the extracted radiomics features through an SVM-RFE method, finally carrying out feature fusion on the radiomics features and the deep learning features through a maximum correlation fusion algorithm, and carrying out classification by using a neural network. According to the technical scheme, deep learning features and radiomics features are combined, and the accuracy of bone mineral density classification is effectively improved.

Description

technical field [0001] The invention relates to the technical field of medical image segmentation and image classification, in particular to a vertebral body bone density classification method based on the fusion of radiomics and deep learning features. Background technique [0002] Osteoporosis is a systemic disease characterized by decreased bone mass, destruction of bone tissue microarchitecture, increased bone fragility, and susceptibility to fractures. Its incidence increases significantly with age. The direct reason is that the detection of bone mass is the key to the prevention and treatment of osteoporosis. The lumbar spine is mostly selected as the measurement site because of its large bone content and uniform distribution, and the comprehensive bone mineral density of multiple vertebral bodies is used for the final evaluation. Standard, usually L1, L2 vertebral bodies. [0003] At present, the commonly used clinical methods for evaluating osteoporosis include dual...

AI Technical Summary

Problems solved by technology

In addition, the existing bone density classification method avoids the comprehensive use of multiple vertebral bodies for evaluation, and the prediction result of a sing

Images