目的 提出一种基于邻域隶属度参数优化的模糊C均值（Fuzzy C-Means, FCM）聚类改进算法，并用于脑部MR图像 分割。方法 首先，采用遗传算法、粒子群优化算法和联合算法计算隶属度函数最佳参数，然后采用此隶属度函数优化FCM 聚类算法相似度函数，最后根据改进的FCM聚类算法分割脑部MR图像。图像分割精确性评价指标采用假阴性率、假阳性 率、分割错误率和相似性系数。结果 选用不同FCM算法对包含噪声的人工合成图像和临床实例MR图像进行仿真实验。定 性分析显示本文提出的FCM聚类改进算法获得分割图像能保留更多的边缘和细节信息；定量评价显示基于本文提出的FCM 改进算法获得的分割假阴性率（0.0058%～4.28%）、假阳性率（0.0182%～20.15%）和错误率（0.0085%～7.15%）均最 小，相似性系数高达92.65%。结论 联合使用遗传算法和粒子群优化算法能获得最佳隶属度参数，基于此改进的FCM聚类算 法能有效克服噪声造成的局限性，提高脑部MR图像分割精准度，具有较高的临床应用价值。

Objective An improved Fuzzy C-means (FCM) clustering algorithm based on neighborhood membership parameter optimization was proposed for MR image segmentation in the brain. Methods First, genetic algorithms (GAs) and particle swarm optimization (PSO) were combined to determine the optimum value of degree of attraction. Then the similarity function was modified by neighborhood attraction. Finally, according to the improved FCM clustering algorithm, brain MR images were segmented. Image segmentation accuracy was measured by the percentage of negative false, positive false, total false, and similarity index. Results Both simulated and real MR images with noise were evaluated by different FCM-based methods. Qualitative analysis showed that edges and small structures were well preserved with the proposed method. Quantitative evaluation results showed that the percentage of negative false (0.0058%-4.28%), positive false (0.0182%-20.15%) and total false (0.0085%-7.15%) were all the minimum, and the similarity index (92.65%) was the largest in the all improved FCM clustering algorithms. Conclusion The combination of genetic algorithm and particle swarm optimization algorithm can obtain the optimal membership parameters. The improved FCM clustering algorithm can effectively overcome the limitations caused by noise and improve the accuracy of MR image segmentation in the brain, which has a high clinical application value.