The geometric preference subtype in ASD: identifying a consistent, early-emerging phenomenon through eye tracking

Abstract

Background: The wide range of ability and disability in ASD creates a need for tools that parse the phenotypic heterogeneity into meaningful subtypes. Using eye tracking, our past studies revealed that when presented with social and geometric images, a subset of ASD toddlers preferred viewing geometric images, and these toddlers also had greater symptom severity than ASD toddlers with greater social attention. This study tests whether this "GeoPref test" effect would generalize across different social stimuli.

Methods: Two hundred and twenty-seven toddlers (76 ASD) watched a 90-s video, the Complex Social GeoPref test, of dynamic geometric images paired with social images of children interacting and moving. Proportion of visual fixation time and number of saccades per second to both images were calculated. To allow for cross-paradigm comparisons, a subset of 126 toddlers also participated in the original GeoPref test. Measures of cognitive and social functioning (MSEL, ADOS, VABS) were collected and related to eye tracking data. To examine utility as a diagnostic indicator to detect ASD toddlers, validation statistics (e.g., sensitivity, specificity, ROC, AUC) were calculated for the Complex Social GeoPref test alone and when combined with the original GeoPref test.

Results: ASD toddlers spent a significantly greater amount of time viewing geometric images than any other diagnostic group. Fixation patterns from ASD toddlers who participated in both tests revealed a significant correlation, supporting the idea that these tests identify a phenotypically meaningful ASD subgroup. Combined use of both original and Complex Social GeoPref tests identified a subgroup of about 1 in 3 ASD toddlers from the "GeoPref" subtype (sensitivity 35%, specificity 94%, AUC 0.75.) Replicating our previous studies, more time looking at geometric images was associated with significantly greater ADOS symptom severity.

Conclusions: Regardless of the complexity of the social images used (low in the original GeoPref test vs high in the new Complex Social GeoPref test), eye tracking of toddlers can accurately identify a specific ASD "GeoPref" subtype with elevated symptom severity. The GeoPref tests are predictive of ASD at the individual subject level and thus potentially useful for various clinical applications (e.g., early identification, prognosis, or development of subtype-specific treatments).

Keywords: Autism spectrum disorder; Early identification; Eye tracking; Geometric preference; Social attention.

Fig. 1

Examples of stimuli. a Sample image from the original GeoPref test which consists of a 60-s video composed of 28 scenes, 26 of which involve one individual’s movement (while the other two scenes include twins moving side by side). ©2003 Gaiam Americas, Inc., Courtesy of Gaiam Americas, Inc. As each social scene switches to a new actor, the paired geometric scene switches simultaneously to different colors and moving shapes. b Sample image from the Complex Social GeoPref test which consists of a 90-s video composed of nine scenes, five of which include two children interacting and four of which include one child moving enthusiastically. As each social scene changes, the paired geometric scene also switches simultaneously to a different color and moving shape, as in the original GeoPref test

Fig. 2

Scatterplot of % time viewing geometric images (total fixation duration while viewing the dynamic geometric stimulus, divided by total fixation duration to the geometric and social stimuli combined) for all subjects who completed the Complex Social GeoPref test (n = 227) sorted by diagnostic group. Boxplots show median, range, and first and third quartiles

Fig. 3

Correlation between original and Complex Social GeoPref Tests. Scatterplot illustrating % Geo scores (summed fixation duration while viewing the dynamic geometric stimulus, divided by total fixation duration to the geometric and social stimuli combined) for each test for each subject in the ASD group who completed both eye tracking tests on separate visits across the span of the study