A method for extracting tumor texture features based on multi-scale technology

Abstract

The invention discloses a multi-scale technology-based tumor texture feature extraction method. The method includes the following steps that: tumor PET image data are obtained from clinic, an ideal tumor treatment region is drawn; segmentation is carried out according to the boundary of the drawn target region, and gross tumor volume (GTV) is extracted, and PET images only containing the GTV are obtained; gray level quantization is performed on each PET image, and two-dimensional dual-tree complex wavelet transform is performed on the quantized images, and a plurality of sub-band images can be obtained from each image; all the sub-band images which are obtained by decomposing the PET images are re-combined into new stereo GTV; the mean and variance of a detail coefficient matrix obtained after two-dimensional dual-tree complex wavelet transform are calculated, and tumor texture feature values are calculated; and the relationships among the maximum values (SUVmax) of the tumor texture feature values and standardized uptake values, tumor TNM stages and infiltration depth are established, and the heterogeneous texture feature values of a quantized tumor are screened and optimized.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method for extracting tumor texture features based on multi-scale technology. Background technique [0002] Texture feature is the most basic visual feature of an image, and texture analysis technology has always been a research hotspot in the field of medical image processing. Fluorodeoxyglucose positron tomography (PET) labeled with F-18 has been widely used in predicting the curative effect of radiotherapy and chemotherapy in cancer patients, guiding individual treatment and prognosis. Studies in recent years have shown that the gray texture change characteristics of tumors in PET images can effectively evaluate the level and heterogeneity of FDG uptake in tumors, so that the degree of heterogeneity of tumors can be effectively quantified, and the obtained heterogeneity parameters can be used To predict the efficacy of radiotherapy and chemotherapy or to guide radiotherapy. ...

AI Technical Summary

Problems solved by technology

The traditional pyramidal wavelet transform only decomposes the low-frequency part at each decomposition level, and only uses the information of the low-frequency sub-band of the texture image when extracting texture features, and cannot use the important texture feature information in the high-frequency sub-band

Images