基于克隆选择算法和马尔可夫链蒙特卡尔方法融合隐马尔可夫随机场的脑部核磁共振自动分割研究

Research of Brain MRI Segmentation Based on HMRF Fusion with CSA and MCMC Methods

针对脑部磁共振(MR)图像分割问题,提出了一种基于克隆选择算法(CSA)和隐马尔可夫随机场(HMRF)的融合方法。首先,采用马尔可夫链蒙特卡尔(MCMC)算法对类标签进行估计,进行体素分类;然后,对分割结果进行偏场校正;最后,利用CSA的统计学进行HMRF模型参数估计,并利用迭代优化算法获得最终的分割结果。由于MCMC和CSA都是全局优化技术,所以HMRF-CSA算法能够克服传统HMRF方法的局部收敛以及较低分割精度的缺点。在仿真脑部MR图像集BrainWeb上的实验结果表明,对于主要脑部结构,算法的分割精度高于其他几种算法;且对图像伪影具有鲁棒性。

For the issue thatbrain magnetic resonance （MR） image segmentation,a fusion method based on clonal selection algorithm （CSA） and Hidden Markov random field （HMRF） is proposed.First,using Markov chain Monte Carl （MCMC） algorithm to estimate the class labels carry voxel classification; Then,the segmentation results bias field correction.Finally,the CSA conducted HMRF statistical model parameter estimation,and using an iterative optimization algorithm to obtain the final segmentation result.Since the MCMC and CSA are global optimization techniques,so HMRF-CSA algorithm can overcome the traditional methods HMRF local convergence and lower segmentation accuracy shortcomings.Experimental results on brain MR image sets BrainWeb simulation showed that the main brain structure,segmentation accuracy of the algorithm is higher than several other algorithms for image artifacts and robust.