肝脏磁共振图像分割方法的研究

Study on Segmentation of Liver Magnetic Resonance Imaging

肝脏磁共振图像分割是诊断肝脏疾病的重要手段之一,偏移场是肝脏磁共振图像中通常存在的灰度不均匀现象,导致图像分割效果不理想。采用正则化相邻局部灰度聚类算法对肝脏磁共振图像进行偏移场矫正,在能量函数中加入辅助变量来解决非凸隶属函数,通过迭代计算得到最优偏移场矫正结果,并采用基于水平集的图像分割算法对矫正后图像进行分割。通过差值图像、灰度直方图、Jaccard参数和Dice参数来评价算法的性能及图像分割效果,并与非参数非均匀归一化变量(N4)算法进行对比分析。实验结果表明,相邻局部灰度聚类算法引入总变分项进行偏移场矫正后,图像与原始图像差值更大,灰度分布更均匀,图像分割结果更准确,便于医生进行病理观察及诊断。

Segmentation of liver magnetic resonance(MR) image is one of the most important methods for diagnosis of liver diseases. The bias field is a common phenomenon of uneven grayscale in liver MR images,which leads to suboptimal image segmentation. In this paper, the coherent local intensity clustering(CLIC)algorithm with regularization is used to correct the bias field of liver MR images. The auxiliary variable is added to the energy function to solve the non-convex membership function. The correction result of optimal bias field is achieved by iterative calculation. An image segmentation algorithm based on level set is used to segment the corrected image. The performance of the algorithm and the effect of image segmentation are evaluated by difference image, gray histogram, Jaccard index and Dice index, and then compared with the nonparametric nonuniform normalization variant(N4) algorithm. The experimental results show that the difference between the corrected image and original image is larger, the gray distribution is more uniform, and the segmentation result is more accurate after introducing a total variation term to CLIC algorithm to correct the bias field, which is convenient for doctors to make pathological observation and diagnosis.