Classification of chromatic visual evoked potentials with the aid of a neural net

Abstract

Twelve normal subjects, and six color blind (three protanopes and three deuteranopes) individuals were used in this study. Visual evoked potentials (50-150 msec post-stimulus) were recorded in response to three stimuli: (a) three horizontal, achromatic bars alternating in luminosity between bright and dim, (b) bars alternating between blue and red, and (c) bars alternating between green and red. The resulting waveforms were normalized in amplitude and submitted to a commercial neural net program for classification. The network correctly identified 24 of the 36 normal responses. (2) The network was also asked to distinguish between the responses of normal and color blind individuals. Based upon the blue/red response, the network correctly classed 12 of 18 responses, and based upon the green/red response correctly classified 14 of 18 (including all 6 color blinds). (3) These results are statistically highly significant and suggest that the VEP elicited by chromatic stimulation is substantially different from that elicited by achromatic stimuli.