High-throughput classification of clinical populations from natural viewing eye movements

Abstract

Many high-prevalence neurological disorders involve dysfunctions of oculomotor control and attention, including attention deficit hyperactivity disorder (ADHD), fetal alcohol spectrum disorder (FASD), and Parkinson's disease (PD). Previous studies have examined these deficits with clinical neurological evaluation, structured behavioral tasks, and neuroimaging. Yet, time and monetary costs prevent deploying these evaluations to large at-risk populations, which is critically important for earlier detection and better treatment. We devised a high-throughput, low-cost method where participants simply watched television while we recorded their eye movements. We combined eye-tracking data from patients and controls with a computational model of visual attention to extract 224 quantitative features. Using machine learning in a workflow inspired by microarray analysis, we identified critical features that differentiate patients from control subjects. With eye movement traces recorded from only 15 min of videos, we classified PD versus age-matched controls with 89.6 % accuracy (chance 63.2 %), and ADHD versus FASD versus control children with 77.3 % accuracy (chance 40.4 %). Our technique provides new quantitative insights into which aspects of attention and gaze control are affected by specific disorders. There is considerable promise in using this approach as a potential screening tool that is easily deployed, low-cost, and high-throughput for clinical disorders, especially in young children and elderly populations who may be less compliant to traditional evaluation tests.

Humans have idiosyncratic and task-specific scanpaths for judging faces

“…If this inference could be reliably made, MFPA would offer a low-cost diagnostic technique (especially since some eye trackers can now be purchased for less than $100). This approach has already shown success in predicting schizophrenia, attention deficit hyperactivity disorder, fetal alcohol spectrum disorder, and Parkinson's disease using algorithms similar to the summary statistics method (Benson et al, 2012;Tseng et al, 2013). Fisher vector methods that incorporate the spatio-temporal characteristics of scanpaths directly could lead to further improvements in disease diagnosis from eye movements.…”

“…Multi-fixation pattern analysis (MFPA) is a new eye movement analysis technique that harnesses machine learning to make inferences about subjects from their eye movements (Benson et al, 2012;Greene, Liu, & Wolfe, 2012;Tseng et al, 2013;Kanan et al, 2014;Borji & Itti, 2014). MFPA algorithms take a person's scanpath, a sequence of fixations, as their input and use the scanpath to infer traits such as the task the person was given.…”

“…If an algorithm can make this inference above chance when trained on a person's scanpaths for specific tasks, then this suggests that the person might have scanpath routines for accomplishing one or more of the tasks. Prior work with MFPA has focused on validating the technique for inferring the task given to a subject when viewing scenes (Greene, Liu, & Wolfe, 2012;Kanan et al, 2014;Borji & Itti, 2014) and for inferring whether the subject has a particular disease (Benson et al, 2012;Tseng et al, 2013). In this paper, we use MFPA to determine if people have scanpath routines for making different inferences about faces.…”

Predicting an observer's task using multi-fixation pattern analysis

“…One promising area of study is the identification of neurophysiological diseases. Tseng et al [2012] pioneered this approach, showing that their algorithm could discern whether their subjects had Attention Deficit Hyperactivity Disorder, Parkinson's Disease, Fetal Alcohol Syndrome, or were disease free by combining visual saliency and eye movement data. Our results suggest that using motor activity alone may be sufficient to make these inferences.…”

Neuro-ophthalmology update

“…Tseng et al devised a universally applicable patient screening tool based on automated analysis of natural viewing eye movements [2]. The participants simply had to watch television while their eye movements were recorded with a head-mounted video-based eye tracker.…”