Radiomics feature screening method based on CC-attention mechanism

Abstract

According to the radiomics feature screening method based on the CC-attention mechanism, feature screening is carried out according to the correlation between the image features and the clinical features, and the situation that a current traditional feature extraction method and the clinical features are not closely connected is improved. The method comprises the following steps: firstly, delineating a region of interest of a colorectal cancer liver metastasis patient to obtain radiomics features, then carrying out correlation analysis on the obtained image features and clinical features to obtain a correlation matrix of the image-clinical features, and then analyzing the image-clinical feature matrix by utilizing a CC-attention mechanism, screening out an image feature group with relatively high correlation with the clinical features, so that the purpose of performing feature screening through the correlation between the clinical features and the image features is achieved. By means of the method, the prediction rate of the colorectal cancer liver metastasis microsatellite state is increased, and tumor treatment is better assisted.

Description

technical field [0001] The invention relates to the technical field of computer-aided correlation analysis, in particular to a radiomics feature screening method based on CC-attention mechanism. Background technique [0002] Colorectal cancer (CRC) is the third most common malignancy and the second leading cause of cancer mortality. Twenty percent of colorectal cancer patients already have liver metastases at the time of diagnosis, and up to 50% of patients will develop liver metastases within the first three years. The incidence of metastasis varies with age, sex, and different primary colorectal cancer sites. Poor differentiation, lymph node metastasis, different metastatic organs, and higher carcinoembryonic antigen were positively correlated with these four types of distant metastasis. According to the frequency of microsatellite instability, it can be divided into three types: microsatellite instability-high (MSI-H), microsatellite instability-low (MSI-L) And microsat...

AI Technical Summary

Problems solved by technology

It is very useful when the number of features needs to be reduced, but it is not very friendly to data understanding. Using the traditional feature screening method, because it is not well combined with clinical features, it may cause the screened features to be useful for classification. effect, but it has little significance for clinical application, and some features not screened by LASSO may have greater clinical significance, which is more meaningful for preoperative diagnosis and treatment of diseases

Images