Single-trial analysis of the auditory N100 improves separation of normal and schizophrenia subjects

Abstract

The N100 component of the auditory evoked potential (EP) has been recently used to study sensory gating deficits in schizophrenia subjects compared to normal controls. Previously, we used selective averaging to show phase synchronization differences in brain activity between the two populations. In this study, we employed our recently developed iterative independent component analysis (iICA) procedure to measure single-trial EPs in the context of a double-stimulus paradigm. Using the amplitude and latency of the N100 components of the first and second stimuli responses obtained from iICA and four different classification algorithms we were able to accurately classify subjects with 100% sensitivity and 100% specificity. In contrast, the same amplitude and latency features computed from average EPs provided only 69% classification accuracy, with 63% sensitivity and 75% specificity, respectively. We conclude that inter-trial temporal variability plays a significant role in the well-known sensory gating deficits found in schizophrenia patients.

Single-subject classification of schizophrenia by event-related potentials during selective attention

“…The balanced classification accuracy of almost 80%, which we obtained, outperforms the results of a previous ERP study using averaged data of auditory N1 (69%; Iyer and Zouridakis, 2008). We are not aware of other single-subject classification studies employing ERP data.…”

“…Most studies report N1 amplitude deficits in schizophrenia throughout sensory domains, yet at the same time largely neglect corresponding latency measures (e.g., Turetsky et al, 2008). Among those studies that describe N1 latency, however, decrements in schizophrenia are not uncommon findings (Mulert et al, 2001;Butler et al, 2007;Iyer and Zouridakis, 2008). Assuming an association between amplitude and latency seems plausible, given the correlation between these measures in our sample.…”

Aberrant auditory evoked responses in schizophrenia: Evidence from single-trial analysis

“…Furthermore, such single-trial estimates allowed accurate separation of normal controls from schizophrenia subjects [9].…”

Schizophrenia Assessment Using Single-Trial Analysis of Brain Activity

“…To explore whether the characteristics of the S1 and S2 responses were statistically distinct to allow automatic subject classification, we analyzed the amplitude and latency values of the P50 and N100 components seen in the S1 and S2 responses, when the latter were identified from ensemble averaging or from single-trial iICA processing [48]. Feature ranking found S1 latency to be the best discriminatory feature, followed by S2 latency, while the S1 and S2 amplitudes had no significant contributions to classification accuracy.…”

“…Moreover, when compared with time-frequency wavelet denoising, iICA performed equally well or better over a wide range of experimental conditions and SNRs [47]. The iICA method has been used to analyze EP activity with data obtained at different labs and under different experimental protocols, including a simple N100 auditory task and typical paired stimulus P50 and N100 paradigms [48,49]. The following sections describe the various steps of the iICA method along with application examples on data recorded from normal control subjects and schizophrenia patients.…”