Department of Electrical Engineering, City University of Hong Kong, Hong Kong 999077, China.
Department of Chemistry, City University of Hong Kong, 83 Tat Chee Avenue, Hong Kong 999077, China.
Sensors (Basel). 2020 Aug 26;20(17):4818. doi: 10.3390/s20174818.
This paper introduces a novel model based on support vector machine with radial basis function kernel (RBF-SVM) using time-series features of zebrafish () locomotion exposed to different electromagnetic fields (EMFs) to indicate the corresponding EMF exposure. A group of 14 adult zebrafish was randomly divided into two groups, 7 in each group; the fish of each group have the novel tank test under a sham or real magnetic exposure of 6.78 MHz and about 1 A/m. Their locomotion in the tests was videotaped to convert into the x, y coordinate time-series of the trajectories for reforming time-series matrices according to different time-series lengths. The time-series features of zebrafish locomotion were calculated by the comparative time-series analyzing framework highly comparative time-series analysis (HCTSA), and a limited number of the time-series features that were most relevant to the EMF exposure conditions were selected using the minimum redundancy maximum relevance (mRMR) algorithm for RBF-SVM classification training. Before this, ambient environmental parameters (AEPs) had little effect on the locomotion performance of zebrafish processed by the empirical method, which had been quantitatively verified by regression using another group of 14 adult zebrafish. The results have demonstrated that the purposed model is capable of accurately indicating different EMF exposures. All classification accuracies can be 100%, and the classification precision of several classifiers based on specific parameters and feature sets with specific dimensions can reach higher than 95%. The speculative reason for this result is that the specified EMF has affected the zebrafish neural aspect, which is then reflected in their behaviors. The outcomes of this study have provided a new indication model for EMF exposures and provided a reference for the investigation of the impact of EMF exposure.
本文提出了一种基于支持向量机和径向基函数核（RBF-SVM）的新型模型，该模型使用斑马鱼运动的时间序列特征来指示相应的电磁场暴露。将 14 条成年斑马鱼随机分为两组，每组 7 条；每组的鱼在假暴露或真实磁场暴露（6.78MHz，约 1A/m）下进行新鱼缸测试。对测试中的运动进行录像，将其转换为轨迹的 x、y 坐标时间序列，根据不同的时间序列长度转换为时间序列矩阵。使用高度比较时间序列分析（HCTSA）比较时间序列分析框架计算斑马鱼运动的时间序列特征，并使用最小冗余最大相关性（mRMR）算法选择与电磁场暴露条件最相关的时间序列特征数量，用于 RBF-SVM 分类训练。在此之前，通过回归分析使用另一组 14 条成年斑马鱼定量验证了环境参数（AEPs）对斑马鱼运动性能的影响很小。经验方法处理的斑马鱼。结果表明，所提出的模型能够准确指示不同的电磁场暴露。所有分类准确率均可达 100%，基于特定参数和特征集的几种分类器的分类精度可达到 95%以上。造成这种结果的推测原因是指定的电磁场已经影响了斑马鱼的神经方面，然后反映在它们的行为中。这项研究的结果为电磁场暴露提供了一种新的指示模型，并为电磁场暴露影响的研究提供了参考。
