基于显著性定位和动态自适应区域生长的HRMR图像斑块分割

为了解决HRMR图像斑块边界模糊及灰度不均匀造成的分割难问题，提出了一种结合显著性定位与改进动态自适应区域生长的斑块分割方法。使用非局部均值滤波算法、平滑梯度实现图像去噪与斑块模糊边缘增强；通过显著性检测获得显著图像，采用形态学开重构得到斑块定位图像；利用改进的动态自适应区域生长算法，实现颅内斑块的准确分割。本实验对象为34组脑血管狭窄患者的HRMR图像，通过与专家手动分割结果对比，斑块的平均分割准确度达到90.16%。研究结果表明，本方法不仅能够提高斑块的分割精度，完整地保留颅内斑块的弱边缘信息，同时还可以避免不同医生手动分割造成的主观差异性，或可用于辅助脑血管狭窄患者的临床诊断与治疗。

Plaque segmentation of HRMR image based on saliency location and dynamic adaptive region growing

In order to solve the problem of segmentation difficulty caused by the fuzzy boundary of plaque and intensity inhomogeneity of HRMR image, this paper proposed a plaque segmentation method combining saliency localization and improved dynamic adaptive region growing. This method used non-local mean filtering algorithm and smooth gradient map to achieve image denoising and plaque blur edge enhancement. It used morphological reconstruction to the saliency detection image obtained by saliency detection to obtain plaque localization image. It employed improved dynamics adaptive region growing algorithm to achieve accurate segmentation of intracranial plaques. The experimental subjects were the HRMR images of 34 patients with cerebral vascular stenosis. The average segmentation accuracy of plaques reached 90.16% by comparison with expert manual segmental results. The results show that the method can not only improve the segmentation accuracy of plaque, and completely preserve the weak edge information of the intracranial plaque, but also avoid the subjective difference caused by manual segmentation by different doctors. This algorithm may be used to assist the clinical diagnosis and treatment of patients with cerebral vascular stenosis.