GrowCut-based nasopharyngeal carcinoma primary lesion clinical target region automatic sketching method and system

Abstract

The embodiment of the invention provides a GrowCut-based nasopharyngeal carcinoma primary lesion clinical target region automatic sketching method and system. The method comprises the following steps:S1, carrying out data discrete operation on an image sequence of a template CT, gridding a three-dimensional CT image to obtain a corresponding three-dimensional grid, reading a tumor contour file drawn by a doctor in a training set, and generating a binary image of a tumor primary lesion area; S2, performing preprocessing operation on the read primary lesion tumor data by adopting opening operation, mining a cross-correlation relationship between adjacent tumors, counting the frequency of joint occurrence of adjacent voxels, storing a statistical result, and establishing a database of cross-correlation rules; S3, reading the primary tumor data of a new patient, and continuously updating the intensity value of each discrete grid point; and giving different loss cost thresholds to obtain corresponding clinical radiotherapy target area ranges. The drawing habit of a doctor is learned, the clinical radiotherapy target area drawing rule is obtained, the drawing time of the doctor is saved, and the result is more reliable.

Description

technical field [0001] The present invention relates to the field of predicting tumor growth and simulating tumor spreading area, in particular to a method and system for automatically delineating clinical target areas of nasopharyngeal carcinoma primary lesions based on GrowCut. Background technique [0002] The existing GrowCut-based automatic clinical target area delineation method for nasopharyngeal carcinoma primary lesions (CN 107437249) includes the following steps: Step S1: discretize the template CT image sequence, grid the three-dimensional CT images to obtain the corresponding three-dimensional network grid, read the tumor outline file drawn by the doctor in the training set, and generate a binary image of the primary tumor area. Step S2: Preprocessing the read-in primary tumor data, mining the correlation between adjacent tumors, counting the frequency of joint occurrence of adjacent voxels, storing the statistical results, and establishing a database of correlat...

AI Technical Summary

Problems solved by technology

During the overall calculation, the probability of internal voxel points may be affected by external points, resulting in a small final result, which does not conform to the actual clinical results.

Images