Pulmonary nodule segmentation method of LBF active contour model based on information entropy and combination vector

Abstract

The invention discloses a pulmonary nodule segmentation method of an LBF active contour model based on information entropy and combination vector. The pulmonary nodule segmentation method is advantageous in that by fully combining with various characteristic information of medical PET images and medical CT images, the SUV values of the PET images are used to acquire a pulmonary nodule region of interest; an initial contour of a nodule is constructed by adopting an automatic threshold value iteration method; a guiding function of a nodule edge evolution is constructed according to the SUV information of the PET images, and by combining PET gray scale combination vectors with CT gray scale combination vectors, the energy functional of the LBF model is improved, and then evolution of a contour curve is accurately stopped at the edge of the pulmonary nodule. The pulmonary nodule segmentation method is advantageous in that operation is simple, and the batch type automatic segmentation of the angiosynizesis type pulmonary nodule is realized, and strong stability and strong accuracy are provided.

Description

technical field [0001] The invention belongs to the field of medical image processing, and specifically relates to a pulmonary nodule segmentation method based on information entropy and a joint vector LBF active contour model, based on various information features of medical PET and CT images, for pulmonary nodules of vascular adhesion type method of segmentation. Background technique [0002] Computer-aided diagnosis (Computer Aided Diagnosis, CAD) system is a new technology that can assist doctors to interpret medical images. Automated second opinions. However, in the diagnosis of pulmonary nodules, due to the complex structure of the lung, the shape and location of the nodules are different in different cases, it is difficult to accurately locate the nodules in the cross-sectional scan of the whole lung by the method of naked eye reading alone. Location and lesion type. In the lung cavity, red blood cells in the blood are rich in nutrients, which provide a good metabo...

AI Technical Summary

Problems solved by technology

It is prone to over-segmentation or under-segmentation during segmentation

If the correct segmentation of nodules cannot be achieved, the subsequent CAD diagnosis process will inevitably be affected

Usually, the segmentation of pulmonary nodules is mostly based on a single CT image. However, this method cannot achieve accurate segmentation of vascular adhesion pulmonary nodules.

The reasons are as follows: (1) Since the gray values ​​of blood vessels and nodules in CT slices are relatively close, it is extremely difficult to effectively separate them, and it is also easy to cause mis-segmentation; (2) In the lung tomography In the picture, the cross-section of blood vessels and nodules are spherical in shape, which is difficult to distinguish

In view of the above problems, only relying on CT images cannot be accurately segmented

Images