基于免疫克隆选择算法搜索GMM的脑岛功能划分

为了得到更好的脑岛功能划分结构，加深人们对其功能组织性的理解，提出一种基于免疫克隆选择（ICS）算法搜索高斯混合模型（GMM）的脑岛功能划分方法（NICS-GMM）.该方法基于功能磁共振成像（fMRI）数据，将GMM映射到抗体上；利用ICS算法搜索能够反映脑岛功能分布的GMM，并在搜索过程中融入具有抗噪能力的动态邻域信息，以提高其搜索质量；利用最优的GMM实现对脑岛的功能划分.在划分数为2~12的脑岛功能划分上，新方法搜得的GMM具有最高的似然分数，而且相应划分结果的轮廓系数也达到了最大值.真实脑岛fMRI数据上的实验结果表明，该方法不仅具有更强的全局搜索能力，还可以得到具有较高功能一致性与更强区域连续性的脑岛功能划分结构.
Abstract: An insula functional parcellation method based on Gaussian mixture model (GMM) searched by immune clonal selection (ICS) algorithm, called NICS-GMM, was presented to get better functional parcellation structure of insula and deepen our understanding of its functional organization. Based on functional magnetic resonance imaging (fMRI) data, the proposed method first mapped a GMM onto an antibody; then ICS algorithm was performed to search a GMM that could reflect insula functional distribution. Meanwhile, dynamic neighborhood information with the anti-noise capability was integrated into the search process to improve search quality of ICS. Finally, insula functional parcellation was obtained by using GMMs with the highest lilelihood scores. The experiments were conducted on real fMRI data of insula with parcellation numbers of 2 to 12. As a result, GMMs obtained by NICS-GMM have the heighest likelyhood scores and the silhouette index values of the corresponding parcellateion results also reach the maximum. The experimental results demonstrate that the proposed method not only has better global search capability, but also can obtain functional parcellation structures of insula with higher functional consistency and stronger regional continuity.