Purchase this article with an account.
Po-He Tseng, Ian Cameron, Doug Munoz, Laurent Itti; Differentiating Patients from Controls by Gazing Patterns. Journal of Vision 2010;10(7):277. doi: 10.1167/10.7.277.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Dysfunction in inhibitory control of attention was shown in children with Attention Deficit Hyperactivity Disorder (ADHD), Fetal Alcohol Spectrum Disorder (FASD), and elderly with Parkinson's Disease (PD). Previous studies explored the deficits in top-down (goal oriented) and bottom-up (stimulus driven) attention with a series of visual tasks. This study investigates the difference in attentional selection mechanism while patients freely viewed natural scene videos without performing specific tasks, and the difference is utilized to develop classifiers to differentiate patients from controls. These specially designed videos are composed of short (2–4 seconds), unrelated clips to reduce top-down expectation and emphasize the difference in gaze allocation at every scene change. Gaze of six groups of observers (control children, ADHD children, FASD children, control young adults, control elderly, and PD elderly) were tracked while they watched the videos. A computational saliency model computed bottom-up saliency maps for each video frame. Correlation between salience and gaze of each population was computed and served as features for classifiers. Leave-one-out was used to train and test the classifiers. With eye traces of less than 4 minutes of videos, the classifier differentiates ADHD, FASD, and control children with 84% accuracy; another classifier differentiates PD and control elderly with 97% accuracy. A feature selection method was also used to identify the features that differentiate the populations the most. Moreover, videos with higher inter-observer variability in gaze were more useful in differentiating populations. This study demonstrates attentional selection mechanisms are influenced by PD, ADHD, and FASD, and the behavioral difference is captured by the correlation between salience and gaze. Furthermore, this task-free method shows promise toward future screening tools.
This PDF is available to Subscribers Only