Abstract
Abstract: :
Purpose: We compared multifocal responses obtained to conventional contrast reversing (CR) and temporally sparse ternary stimuli. A Hierarchical Decomposition[1] (HD) allowed the responses to be decomposed into components in feed-forward and feed back relationships, permitting the HD components to be characterised as being more related to cortical inputs or to cortical processing. Methods: Stimuli containing 8 cortically scaled checkerboards/eye were presented dichoptically at 50.5 frames/eye/s. For CR stimuli each region had contrast –1 or 1 with probability ½. For the sparse stimuli the contrasts are {-1, 0, 1}, where 0 indicates a blank region at the mean luminance, and the transient stimuli were delivered to each region at 6 stimuli/eye/s. The HD method computes Principal Components (PC) and then effects a rotation that satisfies an autoregressive (AR) model. The AR model indicates how much the HD components contribute to themselves, or other the components, at several temporal lags, thus characterising feed -forward and –back relationships. Responses from 46 subjects were further selected for HD analysis on the basis of median signal to noise rations (SNR), or median communalities. Results: Models were computed for each of the 16 stimulus regions, and using a 95% c.l., AR models having 2 lags (10 ms each) were routinely fitted. The proportion of variance accounted for was higher for sparse stimuli, reflecting the better SNRs obtained for sparse (5.32 ± 0.84) vs. CR (3.31 ± 0.53se) stimuli in the 46 subjects. HD analysis showed that the first component provided substantial hierarchical drive to the higher components but little feedback was apparent. Conclusions: The hierarchical relationships indicated that the first component is most likely influenced by cortical input. Higher components had more consistent shapes across visual field regions, possibly indicating that the VEP is better able to characterise cortically intrinsic signals than cortical inputs. The ratio of hierarchical / direct drives was greatest in nasal and inferior visual fields (p=0.05). Superior-peripheral regions had the weakest responses, however, and so were more difficult to characterise. From the perspective of multifocal VEPs as a tool for perimetry the better SNR indicates that sparse stimuli would require about 2.6 × less recording time to achieve the same level of accuracy. The mean recording time here of 3.97 min, or 2 min/eye, was equal for both stimulus conditions. [1]Repucci MA et al. (2001) Ann Biomed Eng 29:1135-49.
Keywords: electrophysiology: non-clinical • electrophysiology: clinical • perimetry