[Maximum entropy method and Fourier transform for wave analysis of pattern visual evoked potentials]

To estimate spectral analysis of pattern visual evoked potentials (P-VEPs) the maximum entropy method (MEM) and fast Fourier transform (FFT) were investigated. P-VEPs were recorded in 16 normal eyes using the reversal pattern of 2 c/d vertical sine wave gratings. Six temporal frequencies (2, 3, 4, 6, 8, 12 Hz) were used. MEM power spectrum with the parameter (dimension of autoregression model) which was decided final prediction error (FPE) method did not necessarily always give satisfactory results. The MEM power spectrum of transient VEP changed in manner different from that of steady state VEP as the dimension of autoregression model varied. In steady state VEP the peak of the MEM spectrum was positioned accurately at dimension 50 but in transient VEP the peak positioned accurately at a larger dimension. Both in MEM and FFT the curves of the square root of the power vs temporal frequency resembled each other and had temporal tuning with a maximum at 6 Hz. At lower temporal frequencies (2, 3 Hz) the correlation coefficient increased gradually as the harmonic component became higher (2 nd less than 4 th less than 6 h). MEM powers in the second harmonic component were highly correlated with FFT power in temporal frequency above 4Hz. These results showed that the MEM power spectrum had higher resolution than FFT and had different properties especially at lower temporal frequencies. MEM with optimum parameters should be very useful for wave analysis of P-VEPs.