Application of optimum linear filter theory to the detection of cortical signals preceding facial movement in cat

BCI Accuracy Using Classifier-Based Latency Estimation and the Optimal Interstimulus Interval

PURPOSE

Detection of event-related potentials (ERPs) in brain-computer interfaces (BCIs) allow for communication by individuals with neuromuscular disorders. Enhancing BCI accuracy may be possible through the exploration of the optimal interstimulus interval (ISI). Our objective is to investigate the relationship between BCI accuracy and the optimal ISI value for an individual.

APPROACH

Using the previously developed classifier-based latency estimation (CBLE) [1], we investigated the relationship between the interstimulus interval (ISI) and P3 Speller BCI accuracy. Participants underwent two consecutive sessions in one day. The first session had a default ISI value of 120ms. An optimal ISI value calculated from the first session was used in the second.

RESULTS

Ten subjects participated in the study. Of the ten, half received an optimal ISI value of 120ms and half 160ms. Accuracy differences after implementing the adjusted ISI ranged from -26.1 percent to 4.35 percent. Suggestions for additional experimental design adjustments are highlighted under the discussion portion of this manuscript.

Latency correction of event-related potentials between different experimental protocols

OBJECTIVE

A fundamental issue in EEG event-related potentials (ERPs) studies is the amount of data required to have an accurate ERP model. This also impacts the time required to train a classifier for a brain-computer interface (BCI). This issue is mainly due to the poor signal-to-noise ratio and the large fluctuations of the EEG caused by several sources of variability. One of these sources is directly related to the experimental protocol or application designed, and may affect the amplitude or latency of ERPs. This usually prevents BCI classifiers from generalizing among different experimental protocols. In this paper, we analyze the effect of the amplitude and the latency variations among different experimental protocols based on the same type of ERP.

APPROACH

We present a method to analyze and compensate for the latency variations in BCI applications. The algorithm has been tested on two widely used ERPs (P300 and observation error potentials), in three experimental protocols in each case. We report the ERP analysis and single-trial classification.

MAIN RESULTS

The results obtained show that the designed experimental protocols significantly affect the latency of the recorded potentials but not the amplitudes.

SIGNIFICANCE

These results show how the use of latency-corrected data can be used to generalize the BCIs, reducing the calibration time when facing a new experimental protocol.

Classifier-based latency estimation: a novel way to estimate and predict BCI accuracy

OBJECTIVE

Brain-computer interfaces (BCIs) that detect event-related potentials (ERPs) rely on classification schemes that are vulnerable to latency jitter, a phenomenon known to occur with ERPs such as the P300 response. The objective of this work was to investigate the role that latency jitter plays in BCI classification.

APPROACH

We developed a novel method, classifier-based latency estimation (CBLE), based on a generalization of Woody filtering. The technique works by presenting the time-shifted data to the classifier, and using the time shift that corresponds to the maximal classifier score.

MAIN RESULTS

The variance of CBLE estimates correlates significantly (p < 10(-42)) with BCI accuracy in the Farwell-Donchin BCI paradigm. Additionally, CBLE predicts same-day accuracy, even from small datasets or datasets that have already been used for classifier training, better than the accuracy on the small dataset (p < 0.05). The technique should be relatively classifier-independent, and the results were confirmed on two linear classifiers.

SIGNIFICANCE

The results suggest that latency jitter may be an important cause of poor BCI performance, and methods that correct for latency jitter may improve that performance. CBLE can also be used to decrease the amount of data needed for accuracy estimation, allowing research on effects with shorter timescales.

Detection of event-related potentials for development of a direct brain interface

The study presented here is part of an ongoing effort to develop a direct brain interface based on detection of event-related potentials (ERPs). In a study presented in a companion article, averaged ERP templates were identified from electrocorticograms recorded during repetition of voluntary motor actions. Here the authors report on the detection of individual motor ERPs within the electrocorticogram using cross-correlation. An averaged ERP template was created from the first half of each electrocorticogram and then cross-correlated with the continuous electrocorticogram from the second half. Points where the cross-correlation value exceeded an experimentally determined detection threshold were considered to be detection points. A detection point was considered to be a valid "hit" if it occurred between 1 second before and 0.25 second after the recorded time of a voluntary action. The difference between the hit and false-positive percentages (HF-difference) was used as a metric of detection accuracy. HF-differences greater than 90 were found for 5 of 15 subjects, HF-differences greater than 75 were found for 8 of 15 subjects, and HF-differences greater than 50 were found for 12 of 15 subjects. The three other subjects with HF-differences less than 50 had electrode locations not well suited for recording movement-related ERPs. Recordings from sensorimotor and supplementary motor areas produced the highest yield of channels with HF-difference greater than 50; however, a number of channels with good performance were found in other areas as well. The results demonstrate the likely prospect of using ERP detection as the basis of a single-switch direct brain interface and that furthermore, there is a good possibility of obtaining multiple control channels using this approach.

Real-time reconstruction of evoked potentials using a new two-dimensional filter method

Standard techniques for evoked potential recording extract a stimulus-locked event from accompanying noise by averaging a large number of sequentially obtained responses. This approach is valid only to the extent that the nervous system's electrical response to successive stimuli is identical. The technique is suboptimal for recording unstable evoked potentials which vary with the subject's state and attention. Similarly, standard methods are suboptimal for efficiently analyzing rapid changes such as may be seen in the operating room. We developed an evoked potential recording method that reconstructs the individual evoked responses (or small subaverages of evoked potentials) to a series of stimuli. First, the raw data from an entire series of one to several hundred responses are recorded digitally. Using a frequency domain two-dimensional filter, the data are then filtered twice, once along the data sequence axis for each trial, and again along the cross-trial sequence axis for comparable frequency coefficients in sequential trials. The reconstructed filtered evoked potentials are plotted, with successive responses stacked for easy tracking of component changes.

Selective averaging of the intracerebral click evoked responses in man: an improved method of measuring latencies and amplitudes

Individual click evoked responses recorded in the right and left frontal lobe in man have been correlated with different templates, that is, particular wave shapes of evoked responses, and from the maxima of the cross correlation functions individual estimates were obtained of amplitude and latency. Averages were then made of evoked responses having particular amplitude and/or latency values. Using such selected trials for averaging, better results were obtained and artificial amplitude damping avoided. It was found for example that the component N90 varied in latency between 75 and 100 msec and was linked with the behaviour of the contralateral recorded potential amplitude of P150. Furthermore it was found that the biphasic potential N230 P430 was very unstable in latency and the amplitude of the averaged potential using all trials was about 30% too small.

Detection of neuroelectric signals from multiple data channels by optimum linear filter methods

A general mathematical formulation for predicting achievable levels of detection of neuroelectric signals in associated background noise is provided for the case where such data are obtainable from multiple recording loci. The formulation depends upon the signal-to-noise ratios and bandwidths of the incorporated data channels and on the degree of noise dependency between channels. The signals are assumed to relate to the same event such as the production of an incipient movement but need not have the same waveshape. The detection technique is based upon passage of the data through a series of optimum linear filters. The outputs of the filters can either be summated in analog fashion prior to making the detection decision, or their separate outputs and separate detection decisions can be treated combinatorially to determine a detection decision for the aggregate. The former method is superior to the latter for small numbers of data channels. The latter method may be preferable where variation in signal latency exists between channels. Incorporation of information from multiple channels with independent noise can result in significant improvement over detection signals from a single channel provided that the signal-to-noise level of each additional channel exceeds that of the aggregate divided by square rootK, K being the total number of added channels. However, the presence of noise dependency between channels may severly restrict the degree of imporvement realizable through the multiple channel detection operation, irrespective of the number of added cha-nels. The implication of this result on the possibility of using EEG signals predicting incipient movement to control the operation of a motor prosthesis is profound. Inter-channel noise dependency with correlation coefficienr filter method of detection levels required for prosthesis operation. Zero lag correlation coefficients between electrical recordings from separate cortical loci both in man a