Liver cancer image classification method and device combining computer vision features and radiomics features

Abstract

The invention discloses a liver cancer image classification method and device combining computer vision features and radiomics features. The method comprises the following steps: 1) collecting a patient clinical image meeting the standard, and preprocessing the collected image; 2) performing computer vision feature extraction on the segmented liver tumor region image; 3) extracting image omics manual features from the segmented liver tumor region images; 4) in combination with computer vision features and radiomics features, performing univariate filtering type screening, and then performing LASSO regression screening; and 5) modeling by using the screened features and clinical features through a multiple logistic regression model, and searching backwards by using a red pool information criterion to select the clinical features suitable for the optimal model, thereby realizing prediction of liver cancer pathology grading. According to the method, image information of more dimensions and levels is considered, the advantages of non-invasion, safety and stability of radiomics are kept, and the method is expected to become an effective clinical preoperative evaluation tool for liver cancer.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a non-invasive preoperative liver cancer pathological grading method combining computer vision features and radiomics features. Background technique [0002] Liver cancer is one of the important causes of tumor morbidity and death in the world, and China accounts for 50% of the world's new liver cancer cases. It endangers human life and health and causes a heavy economic burden to families and society. In the treatment of liver cancer, the degree of differentiation is an important factor affecting the prognosis of patients and the choice of liver transplantation strategy. The pathological grade is a means to express the degree of differentiation of liver cancer. Traditionally, the pathological grade of liver cancer can be obtained through tissue biopsy. However, the accuracy of tumor grading in tumor specimens obtained by biopsy is controversial, and...

AI Technical Summary

Problems solved by technology

However, radiomics features are mostly manual features, which are calculated by shallow mathematical formulas and are easily affected by noise and low-level image features.

Images