Luts Heleen, Van Dun Bram, Alaerts Jane, Wouters Jan
Department of Neurosciences, ExpORL, K.U. Leuven, O & N 2, Leuven, Belgium.
Ear Hear. 2008 Aug;29(4):638-50. doi: 10.1097/AUD.0b013e318174f051.
The purpose of this study was to evaluate the risks of uncontrolled use of an objective detection criterion in recording auditory steady-state responses (ASSRs). The influence of decisions such as when to accept a response and stop the recording was assessed by analyzing the number of false- and true-detected responses.
A large sample of 500 multiple-stimulus ASSR recordings of normal-hearing and hearing-impaired adults and babies was processed offline. Three types of detection paradigms were evaluated. A first type had a fixed recording length with significance testing after the last sweep. A second type allowed a variable recording length and implied sequential application of the statistical decision criterion. The recording was stopped after significance was reached for y consecutive sweeps. The third type was analogous to the second, with the additional requirement of a minimum of eight recorded sweeps. Furthermore, the effect of significance level and averaging procedure were assessed. Error rates were calculated for the different detection paradigms at eight control frequencies. At the signal frequencies, detection rates and recording times were determined, keeping the error rates fixed. Moreover, ASSR thresholds were compared for a selection of detection paradigms.
When a variable recording length was allowed and a significance level of p = 0.05 was applied, the error rate increased to unacceptable levels because of the effect of repeated testing. The error rate decreased as the required number of consecutive significant sweeps increased and approximated 5% only when eight consecutive significant sweeps were required (with a maximum of 32 recorded sweeps). With an error rate of 5%, the highest detection rate was associated with a fixed recording length of 32 sweeps combined with weighted averaging. A substantial decrease in detection rate was noted when less than 24 sweeps were recorded per intensity. All paradigms with a variable recording length had rather comparable detection rates and recording times. With an error rate of only 1%, small responses could not be distinguished from the noise. The reduction in recording time using a variable instead of a fixed recording length was very limited when a conventional multiple-stimulus approach was used. Test duration would be reduced considerably when the test set-up would allow an independent presentation and recording of the eight signals and responses. Differences in overall detection rate had a small effect on the ASSR thresholds.
The error rate, detection rate, and recording time can improve or deteriorate significantly, even with small adaptations of the detection protocol. When a variable recording length is allowed, the acceptance criterion of the statistical test needs to be adjusted to ensure a tolerable error rate. However, most commercial devices do not offer this option. Then, it is advisable to use a fixed recording length and to judge the significance of the responses at the end of the recording. Although response detection is objective, the measurement protocol has to be well-considered and a critical approach is required when interpreting the responses.
本研究旨在评估在记录听觉稳态反应（ASSR）时无节制使用客观检测标准的风险。通过分析错误检测和正确检测反应的数量，评估诸如何时接受反应并停止记录等决策的影响。
对500例正常听力和听力受损的成人及婴儿的多刺激ASSR记录样本进行离线处理。评估了三种检测范式。第一种范式具有固定的记录长度，在最后一次扫描后进行显著性检验。第二种范式允许记录长度可变，并意味着依次应用统计决策标准。在连续y次扫描达到显著性后停止记录。第三种范式与第二种类似，但额外要求至少记录8次扫描。此外，评估了显著性水平和平均程序的影响。计算了八种控制频率下不同检测范式的错误率。在信号频率下，保持错误率固定，确定检测率和记录时间。此外，比较了所选检测范式的ASSR阈值。
当允许记录长度可变且应用p = 0.05的显著性水平时，由于重复测试的影响，错误率增加到不可接受的水平。随着所需连续显著扫描次数的增加，错误率降低，只有当需要连续8次显著扫描（最多记录32次扫描）时，错误率才接近5%。错误率为5%时，最高检测率与32次扫描的固定记录长度和加权平均相结合相关。当每个强度记录的扫描次数少于24次时，检测率显著下降。所有记录长度可变的范式的检测率和记录时间相当。错误率仅为1%时，小反应无法与噪声区分开来。当使用传统的多刺激方法时，使用可变记录长度而非固定记录长度减少的记录时间非常有限。当测试设置允许独立呈现和记录八个信号及反应时，测试持续时间将大幅缩短。总体检测率的差异对ASSR阈值影响较小。
即使检测方案有微小调整，错误率、检测率和记录时间也可能显著改善或恶化。当允许记录长度可变时，需要调整统计检验的接受标准以确保可接受的错误率。然而，大多数商业设备不提供此选项。此时，建议使用固定记录长度，并在记录结束时判断反应的显著性。尽管反应检测是客观的，但测量方案必须经过充分考虑，在解释反应时需要采取批判性方法。
