Predicting Optimal Stopping Criterion for Automatic Detection Using Magnitude-Squared Coherence: Application to EEG During Auditory Stimulation
AbstractDetection of Auditory Steady-State Responses (ASSR) has been used in the estimation of comprehensive physiological audiometric profiles to help patients with cognitive problems. The presence or absence of response can be inferred using techniques for Objective Response Detection (ORD). In clinical tests, where the minimum test time is required, ORD techniques can be applied while the relevant signals are being collected. However, the repetitive use of such techniques can increase false positive rates. An alternative solution to this problem is to define the presence of an ASSR only when the ORD technique detects the presence of response for a number of consecutive tests (NCT). The aim of this study is, therefore, to determine the value of NCT, with a significance level of 5%, as function of the parameters of the detection protocol used in the experiment. To this end, the detection technique considered in this work is the Magnitude-Squared Coherence (MSC) together with the Monte Carlo Method (30,000 replications). The results of applying such a technique shows that the value of NCT can be estimated using a simple linear regression equation with adjusted coefficient of determination (R2 adjusted) of 0.9651. To demonstrate the usefulness of the proposed methodology, the value of NCT is estimated for a set of EEG signals for different protocols and the results show that the false alarm rates are less than 5%.