Reduction of the ERP Measurement Time by a Weighted Averaging Using Deep Learning.

In clinical examination, event-related potentials (ERPs) are estimated by averaging across multiple responses, which suppresses background EEG. However, acquiring the number of responses needed for this process is time consuming. We therefore propose a method for shortening the measurement time using weighted-average processing based on the output of deep learning. Using P300 as a representative component, here we focused on the shape of the ERP and evaluated whether our method emphasizes the P300 peak amplitude more than conventional averaging, while still maintaining the waveform shape and the P300 peak latency. Thus, using either CNN or EEGNet, the correlation coefficient reflecting the waveform shape, the peak P300 amplitude, and the peak latency were evaluated and compared with the same factors obtained from conventional waveform averaging. Additionally, the degree of background EEG suppression provided by our method was evaluated using the root mean square of the pre-stimulation waveform, and the number of fewer responses required for averaging (i.e., the reduction in measurement time) was calculated.The results showed that compared with P300 values obtained through conventional averaging, our method allowed for the same shape and response latency, but with a higher amplitude, while requiring a smaller number of responses. Our method showed that by using EEGNet, measurement time could be reduced by 13.7%. This corresponds to approximately a 40-second reduction for every 5 minutes of measurement time.

Fast EEG spike detection via eigenvalue analysis and clustering of spatial amplitude distribution.

OBJECTIVE: In the current study, we tested a proposed method for fast spike detection in electroencephalography (EEG). APPROACH: We performed eigenvalue analysis in two-dimensional space spanned by gradients calculated from two neighboring samples to detect high-amplitude negative peaks. We extracted the spike candidates by imposing restrictions on parameters regarding spike shape and eigenvalues reflecting detection characteristics of individual medical doctors. We subsequently performed clustering, classifying detected peaks by considering the amplitude distribution at 19 scalp electrodes. Clusters with a small number of candidates were excluded. We then defined a score for eliminating spike candidates for which the pattern of detected electrodes differed from the overall pattern in a cluster. Spikes were detected by setting the score threshold. MAIN RESULTS: Based on visual inspection by a psychiatrist experienced in EEG, we evaluated the proposed method using two statistical measures of precision and recall with respect to detection performance. We found that precision and recall exhibited a trade-off relationship. The average recall value was 0.708 in eight subjects with the score threshold that maximized the F-measure, with 58.6 ± 36.2 spikes per subject. Under this condition, the average precision was 0.390, corresponding to a false positive rate 2.09 times higher than the true positive rate. Analysis of the required processing time revealed that, using a general-purpose computer, our method could be used to perform spike detection in 12.1% of the recording time. The process of narrowing down spike candidates based on shape occupied most of the processing time. SIGNIFICANCE: Although the average recall value was comparable with that of other studies, the proposed method significantly shortened the processing time.

Fast spike detection in EEG using eigenvalue analysis and clustering of spatial amplitude distribution.

In the current study, we tested a proposed method for fast spike detection using a general-purpose computer. First, we performed eigenvalue analysis using a gradient calculated from two neighboring samples to detect high-amplitude negative peaks. Clustering was performed to classify detected peaks by considering amplitude distribution at scalp electrodes. Negative peaks were scored by considering electrodes in the detection process and the cluster to which each peak belonged. Spikes were detected using two parameters: score threshold, and the number of clusters. We then used precision and recall to eliminate overestimation of the performance of the method. The results revealed a tradeoff between precision and recall. Recall showed a maximum average value of 0.90 in two subjects. In contrast, average precision was 0.21, and the false positive rate was almost four times higher than the true positive rate on the condition that 64 and 54 spikes were included in two subjects. Analysis of required processing time revealed that our method could complete spike detection in approximately one-eighth of the recording time.

Quantitative evaluation of photic driving response for computer-aided diagnosis.

The aim of our research is the quantification of the photic driving response, a routine electroencephalogram (EEG) examination, for computer-aided diagnosis. It is well known that the EEG responds not only to the fundamental frequency but also to all sub and higher harmonics of a stimulus. In this study, we propose a method for detecting and evaluating responses in screening data for individuals. This method consists of two comparisons based on statistical tests. One is an intraindividual comparison between the EEG at rest and the photic stimulation (PS) response reflecting enhancement and suppression by PS, and the other is a comparison between data from an individual and a distribution of normals reflecting the position of the individual's data in the distribution of normals in the normal database. These tests were evaluated using the Z-value based on the Mann-Whitney U-test. We measured EEGs from 130 normal subjects and 30 patients with any of schizophrenia, dementia and epilepsy. Normal data were divided into two groups, the first consisting of 100 data for database construction and the second of 30 data for test data. Using our method, a prominent statistical peak of the Z-value was recognized even if the harmonics and alpha band overlapped. Moreover, we found a statistical difference between patients and the normal database at diagnostically helpful frequencies such as subharmonics, the fundamental wave, higher harmonics and the alpha frequency band.