Deconvolution and modeling of overlapping visual evoked potentials

Visual Evoked Potentials (VEPs) are brain signals generated in response to visual stimuli which are used in the diagnosis of neuro-ophthalmic disorders and the development of brain computer interfaces (BCI). When a pattern-reversal stimulus is delivered at low rates, the acquired transient VEP response (VEPtr) represents the activation of neural structures in the visual pathway. For stimulation rates above six reversals per second (rps), response overlapping occurs and steady state responses are formed (VEPss). This study investigates VEP adaptation by deconvolving responses to temporally jittered stimuli at high rates. VEPs were obtained from normal subjects stimulated with a LED based pattern reversal stimulator delivering three steady-state (1.63, 8.14 and 13.02 rps) and two low-jitter (8.14 and 13.02 mean rps) sequences. Using the Continuous Loop Averaging Deconvolution (CLAD) method VEPtrs were extracted from the jittered responses. To verify the quality of the deconvolution process, the high-rate isochronic VEPss were compared with a synthetic response estimated with the VEPtr. Consistent VEPs (P100-N135) were obtained from all subjects at all rates. At high rates, latencies were slightly shorter while amplitudes were reduced differentially for P100 (80%) and N135 (20%). Models constructed using VEPtr, reliably predicted 8rps and 13rps VEPss confirming high SNR obtained in BCI applications.