Identifying Reliable Neural Components and P300 Latency using Correlated Component Analysis

Abstract This paper applies Correlated component analysis (CorrCA) to extract the reliable and maximally correlated P300 component. Trials collected over time during an oddball electroencephalogram (EEG) experiment are first segregated into the target and non-target trial cohorts. CorrCA was then applied to both cohorts separately. Event related potential (ERP) image plots of CorrCA components from target trials showed a coherent structure for P300 latency around 400 ms when compared to plots of single electrode (Cz) trials. A difference in peak amplitude of 0.37 micro volts was observed between averaged first CorrCA component and averaged target trials at Cz electrode for able bodied subjects. Inter trial Correlation (ITC) obtained using CorrCA algorithm is applied to detect subjects who performed poorly during the oddball experiment. We also used a forward model considering sixteen channels to identify the spatial patterns of neural activity for both able and disabled subjects. It was observed that disabled subjects illustrated lower neural activation during target trials in most brain areas when compared to able-bodied subjects.