MRCG Algorithm And Its Application In Ultrasound Image Segmentation

HIFU ntensity High ocused Id Ultrasoun F)(is a treatment without incision.It’s gradually used in the treatment of tumors in ultrasound images.In the process of treatment,the segmentation of tumor region is mostly artificial segmentation,and computer-aided segmentation has become a research hotspot.Based on IMRG(Iterative multiple region growing),MRCG(Multiple region ecompetitiv growing)algorithm is proposed,which is used to generate Superpixels.It is quantitatively compared their boundary adherence,segmentation accuracy and running time with state-of-the-art superpixel algorithms.The comprehensive performance of MRCG is better.A self-tuning spectral clustering Algorithm based on MRCG is proposed.The selftuning spectral clustering Algorithm is used to merge the superpixels generated by MRCG and segment the tumor areas in the ultrasound image.The self-tuning spectral clustering algorithm can adapt to the number of clustering in the merging stage and reduce the manual intervention.The average similarity is 87.70% compared with manual results.For a 204×228 image,the running time of the self-tuning spectral clustering algorithm based on MRCG algorithm is 12.316 s.Finally,the MRCG method is improved to make it possible to cluster based on superpixel,and the improved method is applied to the segmentation of ultrasound tumor images.To reduce manual intervention,The overlapping regions of tumor regions segmented by different cluster numbers are used to estimate the regions that may belong to the real tumor regions.Then it is used to segment and screen the best tumor regions to get the optimal number of clusters.For a 204×228 image,the number of clustering numbers is set to 14.The segmentation method based on MRCG takes a total running time of 1.568 s and 0.057 s for the merging stage.The average similarity is 87.96%.The method based on MRCG has a great advantage in the evaluation index of contour.