Williams Nikolas S, McArthur Genevieve M, Badcock Nicholas A
Department of Cognitive Science, Macquarie University, Sydney, NSW, Australia.
School of Psychological Science, University of Western Australia, Perth, WA, Australia.
PeerJ. 2021 Feb 9;9:e10700. doi: 10.7717/peerj.10700. eCollection 2021.
The use of consumer-grade electroencephalography (EEG) systems for research purposes has become more prevalent. In event-related potential (ERP) research, it is critical that these systems have precise and accurate timing. The aim of the current study was to investigate the timing reliability of event-marking solutions used with Emotiv commercial EEG systems.
We conducted three experiments. In Experiment 1 we established a jitter threshold (i.e. the point at which jitter made an event-marking method unreliable). To do this, we introduced statistical noise to the temporal position of event-marks of a pre-existing ERP dataset (recorded with a research-grade system, Neuroscan SynAmps at 1,000 Hz using parallel-port event-marking) and calculated the level at which the waveform peaks differed statistically from the original waveform. In Experiment 2 we established a method to identify 'true' events (i.e. when an event appear in the EEG data). We did this by inserting 1,000 events into Neuroscan data using a custom-built event-marking system, the 'Airmarker', which marks events by triggering voltage spikes in two EEG channels. We used the lag between Airmarker events and events generated by Neuroscan as a reference for comparisons in Experiment 3. In Experiment 3 we measured the precision and accuracy of three types of Emotiv event-marking by generating 1,000 events, 1 s apart. We measured precision as the variability (standard deviation in ms) of Emotiv events and accuracy as the mean difference between Emotiv events and true events. The three triggering methods we tested were: (1) Parallel-port-generated TTL triggers; (2) Arduino-generated TTL triggers; and (3) Serial-port triggers. In Methods 1 and 2 we used an auxiliary device, Emotiv Extender, to incorporate triggers into the EEG data. We tested these event-marking methods across three configurations of Emotiv EEG systems: (1) Emotiv EPOC+ sampling at 128 Hz; (2) Emotiv EPOC+ sampling at 256 Hz; and (3) Emotiv EPOC Flex sampling at 128 Hz.
In Experiment 1 we found that the smaller P1 and N1 peaks were attenuated at lower levels of jitter relative to the larger P2 peak (21 ms, 16 ms, and 45 ms for P1, N1, and P2, respectively). In Experiment 2, we found an average lag of 30.96 ms for Airmarker events relative to Neuroscan events. In Experiment 3, we found some lag in all configurations. However, all configurations exhibited precision of less than a single sample, with serial-port-marking the most precise when paired with EPOC+ sampling at 256 Hz.
All Emotiv event-marking methods and configurations that we tested were precise enough for ERP research as the precision of each method would provide ERP waveforms statistically equivalent to a research-standard system. Though all systems exhibited some level of inaccuracy, researchers could easily account for these during data processing.
将消费级脑电图（EEG）系统用于研究目的已变得更为普遍。在事件相关电位（ERP）研究中，这些系统具备精确准确的计时至关重要。本研究的目的是调查与Emotiv商用EEG系统一起使用的事件标记解决方案的计时可靠性。
我们进行了三项实验。在实验1中，我们确定了抖动阈值（即抖动使事件标记方法不可靠的点）。为此，我们将统计噪声引入预先存在的ERP数据集（使用研究级系统Neuroscan SynAmps以1000 Hz记录，通过并行端口事件标记）的事件标记的时间位置，并计算波形峰值在统计上与原始波形不同的水平。在实验2中，我们建立了一种识别“真实”事件的方法（即事件在EEG数据中出现的时间）。我们通过使用定制的事件标记系统“Airmarker”将1000个事件插入Neuroscan数据中来实现这一点，该系统通过触发两个EEG通道中的电压尖峰来标记事件。我们将Airmarker事件与Neuroscan生成的事件之间的延迟用作实验3中比较的参考。在实验3中，我们通过生成间隔1秒的1000个事件来测量三种类型的Emotiv事件标记的精度和准确性。我们将精度测量为Emotiv事件的变异性（以毫秒为单位的标准差），准确性测量为Emotiv事件与真实事件之间的平均差异。我们测试的三种触发方法是：（1）并行端口生成的TTL触发；（2）Arduino生成的TTL触发；（3）串行端口触发。在方法1和2中，我们使用辅助设备Emotiv Extender将触发信号合并到EEG数据中。我们在Emotiv EEG系统的三种配置中测试了这些事件标记方法：（1）Emotiv EPOC +以128 Hz采样；（2）Emotiv EPOC +以256 Hz采样；（3）Emotiv EPOC Flex以128 Hz采样。
在实验1中，我们发现较小的P1和N1峰值在相对于较大的P2峰值较低的抖动水平下衰减（P1、N1和P2分别为21毫秒、16毫秒和45毫秒）。在实验2中，我们发现Airmarker事件相对于Neuroscan事件的平均延迟为30.96毫秒。在实验3中，我们发现在所有配置中都存在一些延迟。然而，所有配置的精度都小于单个样本，当与256 Hz的EPOC +采样配对时，串行端口标记最精确。
我们测试的所有Emotiv事件标记方法和配置对于ERP研究来说都足够精确，因为每种方法的精度将提供与研究标准系统在统计上等效的ERP波形。虽然所有系统都表现出一定程度的不准确性，但研究人员在数据处理过程中可以轻松考虑到这些因素。
