Raeisi Khadijeh, Mohebbi Maryam, Khazaei Mohammad, Seraji Masoud, Yoonessi Ali
School of Electrical Engineering, K.N.Toosi University of Technology, Tehran, Iran.
School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran.
Comput Biol Med. 2020 Feb;117:103596. doi: 10.1016/j.compbiomed.2019.103596. Epub 2019 Dec 30.
Despite the widespread prevalence of Multiple Sclerosis (MS), the study of brain interactions is still poorly understood. Moreover, there has always been a great need to automate the MS diagnosis procedure to eliminate the evaluation errors thereby improving its consistency and reliability. To address these issues, in this work, we proposed a robust pattern recognition algorithm as a computer-aided diagnosis system. This method is based on calculating the pairwise phase-synchrony of EEG recordings during a visual task. Initially, the bivariate empirical mode decomposition (BEMD) was applied to extract the intrinsic mode functions (IMFs). The phases of these IMFs were then obtained using the Hilbert transform to be utilized in the mean phase coherence (MPC), a measure for phase-synchrony calculation. After the construction of the feature space using MPC values, the ReliefF algorithm was applied for dimension reduction. Finally, the best distinguishing features were input to a k-nearest neighbor (KNN) classifier. The results revealed a higher level of network synchronization in the posterior regions of the brain and desynchronization in the anterior regions among the MS group as compared with the normal subjects. In the validation phase, the leave-one-subject-out cross-validation (LOOCV) method was used to assess the validity of the proposed algorithm. We achieved an accuracy, sensitivity, and specificity of 93.09%, 91.07%, and 95.24% for red-green, 90.44%, 88.39%, and 92.62% for luminance, and 87.44%, 87.05%, and 87.86% for blue-yellow tasks, respectively. The experimental results demonstrated the reliability of the presented method to be generalized in the field of automated MS diagnosis systems.
尽管多发性硬化症（MS）普遍存在，但对大脑相互作用的研究仍了解不足。此外，一直非常需要使MS诊断程序自动化，以消除评估误差，从而提高其一致性和可靠性。为了解决这些问题，在这项工作中，我们提出了一种强大的模式识别算法作为计算机辅助诊断系统。该方法基于计算视觉任务期间脑电图记录的成对相位同步。首先，应用双变量经验模式分解（BEMD）来提取本征模函数（IMF）。然后使用希尔伯特变换获得这些IMF的相位，以用于平均相位相干（MPC），这是一种用于相位同步计算的度量。使用MPC值构建特征空间后，应用ReliefF算法进行降维。最后，将最佳区分特征输入到k近邻（KNN）分类器中。结果显示，与正常受试者相比，MS组大脑后部区域的网络同步水平更高，前部区域出现去同步。在验证阶段，使用留一法交叉验证（LOOCV）方法评估所提出算法的有效性。对于红绿色任务，我们分别实现了93.09%、91.07%和95.24%的准确率、灵敏度和特异性；对于亮度任务，分别为90.44%、88.39%和92.62%；对于蓝黄色任务，分别为87.44%、87.05%和87.86%。实验结果证明了所提出方法在自动化MS诊断系统领域中具有可推广的可靠性。
