Abstract
Abstract: :
Purpose:A completely blind retinis pigmentosa volunteer chronically implanted with an optic nerve prosthesis has learned to discriminate one object among 6 (Lambert et al., ARVO 2003). In this study the recognition strategies of the volunteer were analysed with the aim (1) to verify if she was able to apply the scanning strategy she was asked to follow; (2) to evaluate the frequency at which each of the phosphenes available in the surviving part of the visual field is used; and (3) to attempt to find a consistent pattern of recognition. Methods:The prosthesis was composed of a head worn video camera connected to an external processor which generated the appropriate parameters to be send to an implanted stimulator. The nerve was specifically stimulated whenever a part of the processed image of an examined object matched the center of an elicitable phosphene. The remaining visual field included 119 different phosphenes whose center roughly extended horizontally from x= –7° to 7° and vertically from y = –35° to +6°. The total duration of the recognition time was systematically recorded. After several months of recognition training, additional sessions were performed, in which on–line images of superimposition of elicited phosphenes and the examined object were numerically video–recorded, and then analyzed frame–by–frame. Results:The volunteer systematically performed several left–right and bottom–up scans to recognize each object, according to the strategy required by the examiners. Phosphenes located above y= –7° were less frequently used than those of the lower visual field. Phosphenes whose centers were located at x=4° and 5° and y between –8° and –15° were consistently used more frequently in those trials where best performances were observed (mean recognition time of 23 sec). In addition, recognition strategies appeared to be composed of an exploration phase followed by some speculation time (staring next to the object). Both exploration and staring phases had a similar duration of approximately 3 to 4 s and were repeated a number of times. Conclusions:The volunteer was able to adhere to the experimenters' instructions which is encouraging for rehabilitation procedures. Improved recognition strategies could thus be developed on the basis of the above findings. For example the degree of matching between the stimulation provided and the visual imagery of the volunteer should be optimized to reduce the staring phases.
Keywords: image processing • visual search • shape, form, contour, object perception