Townsend G, Platsko V
Algoma University, Sault Ste. Marie, Ontario, P6A 2G4, Canada.
J Neural Eng. 2016 Apr;13(2):026024. doi: 10.1088/1741-2560/13/2/026024. Epub 2016 Feb 25.
A new presentation paradigm for the P300-based brain-computer interface (BCI) referred to as the 'asynchronous paradigm' (ASP) is introduced and studied. It is based on the principle of performance guided constraints (Townsend et al 2012 Neurosci. Lett. 531 63-8) extended from the spatial domain into the temporal domain. The traditional constraint of flashing targets in predefined constant epochs of time is eliminated and targets flash asynchronously with timing based instead on constraints intended to improve performance.
We propose appropriate temporal constraints to derive the ASP and compare its performance to that of the 'checkerboard paradigm' (CBP), which has previously been shown to be superior to the standard 'row/column paradigm' introduced by Farwell and Donchin (1988 Electroencephalogr. Clin. Neurophysiol. 70 510-23). Ten participants were tested in the ASP and CBP conditions both with traditional flashing items and with flashing faces in place of the targets (see Zhang et al 2012 J. Neural Eng. 9 026018; Kaufmann and Kübler 2014 J. Neural Eng. 11 ; Chen et al 2015 J. Neurosci. Methods 239 18-27). Eleven minutes of calibration data were used as input to a stepwise linear discriminant analysis to derive classification coefficients used for online classification.
Accuracy was consistently high for both paradigms (87% and 93%) while information transfer rate was 45% higher for the ASP than the CBP. In a free spelling task, one subject spelled a 66 character sentence (from a 72 item matrix) with 100% accuracy in 3 min and 24 s demonstrating a practical throughput of 120 bits per minute (bpm) with a theoretical upper bound of 258 bpm. The subject repeated the task three times in a row without error.
This work represents an advance in P300 speller technology and raises the ceiling that was being reached on P300-based BCIs. Most importantly, the research presented here is a novel and effective general strategy for organising timing for flashing items. The ASP is only one possible implementation of this work since in general it can be used to describe all previous existing presentation paradigms as well as any possible new ones. This may be especially important for people with neuromuscular disabilities.
介绍并研究一种基于P300的脑机接口（BCI）的新呈现范式，即“异步范式”（ASP）。它基于性能引导约束原则（汤森等人，2012年，《神经科学快报》，第531卷，第63 - 8页），该原则从空间域扩展到了时间域。消除了在预定义固定时间段内闪烁目标的传统约束，目标改为异步闪烁，其时间安排基于旨在提高性能的约束条件。
我们提出了合适的时间约束来推导ASP，并将其性能与“棋盘范式”（CBP）进行比较，此前已证明CBP优于法威尔和唐钦（1988年，《脑电图与临床神经生理学》，第70卷，第510 - 23页）引入的标准“行/列范式”。10名参与者在ASP和CBP条件下进行测试，测试项目既有传统的闪烁物品，也有用闪烁面部代替目标的情况（见张等人，2012年，《神经工程学杂志》，第9卷，026018；考夫曼和屈布勒，2014年，《神经工程学杂志》，第11卷；陈等人，2015年，《神经科学方法杂志》，第239卷，第18 - 27页）。11分钟的校准数据被用作逐步线性判别分析的输入，以推导用于在线分类的分类系数。
两种范式的准确率都一直很高（分别为87%和93%），而ASP的信息传输率比CBP高45%。在一个自由拼写任务中，一名受试者在3分24秒内以100%的准确率拼出了一个66字符的句子（从一个72项矩阵中选取），实际吞吐量为每分钟12比特（bpm），理论上限为258 bpm。该受试者连续三次重复该任务且无错误。
这项工作代表了P300拼写技术的进步，并提高了基于P300的脑机接口所达到的上限。最重要的是，这里提出的研究是一种新颖且有效的用于组织闪烁项目时间安排的通用策略。ASP只是这项工作的一种可能实现方式，因为总体而言它可用于描述所有先前现有的呈现范式以及任何可能的新范式。这对于神经肌肉残疾患者可能尤为重要。
