= PDF Reprint,     = BibTeX entry,     = Online Abstract

Click to download BibTeX data Clik to view abstract L. Itti, M. Yoshida, D. J. Berg, T. Ikeda, R. Kato, K. Takaura, T. Isa, Investigation of spontaneous saccades based on the saliency model in monkeys with unilateral lesion of primary visual cortex, In: Proc. Vision Science Society Annual Meeting (VSS07), May 2007.

Abstract: Primary visual cortex (area V1) is the entry point of visual processing into the primate cortex. Yet, human and animal studies of V1 lesions have demonstrated a ``blindsight'' phenomenon, whereby residual visually-guided behavior remains even when large portions of V1 are absent. However, little is known quantitatively of how this residual vision (possibly subcortically-mediated) differs from normal (cortically-mediated) vision. We analyzed eye movements of three macaque monkeys (one normal, two with complete unilateral V1 ablation) watching ~54 minutes of television (97,051 video frames). A computational model of bottom-up attention quantified how salient visual features may guide gaze into the normal vs. the lesioned hemifield. To eliminate stimulus biases, we randomly presented all video clips twice, original and horizontally flipped. We quantified the extent to which salient stimuli attracted gaze of each monkey by computing, for saccades tallied along the eight principal directions, a model-based bottom-up guidance score (chance level 0.5; ideal upper bound 1.0; practical inter-observer score previously measured as the extent to which three control monkeys predict gaze of a fourth monkey ~0.6). For the normal as well as both lesioned monkeys, we found that saccades in all eight principal directions were guided towards salient locations, significantly above chance (scores 0.585+/-0.006 to 0.662+/-0.004, t-tests p<0.00003, 909 to 3,537 saccades per monkey in each direction). However, although lesioned monkeys overall scored lower, there was little difference in bottom-up guidance with saccade direction (scores 2.2 percent to 3.6 percent lower for saccades directed into vs. away from the lesioned hemifield). Our preliminary results suggest that the extent to which monkey saccades are attracted towards salient locations during natural vision may be less affected by the absence of primary visual cortex than previously considered.

Themes: Computational Modeling, Bayesian Theory of Surprise, Human Eye-Tracking Research, Model of Bottom-Up Saliency-Based Visual Attention


Copyright © 2000-2007 by the University of Southern California, iLab and Prof. Laurent Itti.
This page generated by bibTOhtml on Fri Jan 26 09:25:23 PST 2018