September 2019
Volume 19, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2019
Identifying Scanpath Trends using a Frequent Trajectory Pattern Mining Approach
Author Affiliations & Notes
  • Brian R King
    Department of Computer Science, Bucknell University,
  • Vanessa Troiani
    Geisinger Autism and Developmental Medicine Institute (ADMI), Geisinger Health System
Journal of Vision September 2019, Vol.19, 307a. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Brian R King, Vanessa Troiani; Identifying Scanpath Trends using a Frequent Trajectory Pattern Mining Approach. Journal of Vision 2019;19(10):307a.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Eye tracking systems have a large number of methods available to identify and visualize fixations, saccades, and scanpaths from individual participants in a study. However, methods to extract meaningful, statistically-significant group-level eye movement behaviors from a large cohort of subjects from multiple trials in a study are limited. Heat maps aggregate multiple trials over time, providing the means to visually identify fixations most common among multiple participants. However, heatmap methods discard the temporal view of the data, making it difficult to identify common patterns and trends in eye gaze behavior within subject groups, and even more challenging to compare gaze behavior between groups observing the same stimulus. We present a new method to extract frequent, significant trends in scanpaths from a cohort of participants observing static images. Our method is based on techniques used in the fields of sequential pattern mining and trajectory data mining, which are regularly used in mapping and navigation systems. Similar to how these systems analyze trends from large quantities of geospatial / global positioning system (GPS) data, our method analyzes eye tracking data from multiple subjects to identify frequently occurring scanpath trends. Identified trends are tested for statistical significance using bootstrap sampling. Scanpath trends are visualized, with p-values indicated, providing an easily interpretable plot for establishing the most significant trends among the cohort being evaluated. To validate the method, we implemented preliminary software in Python and evaluated eye tracking data collected from 60 participants observing a 2D static stimulus used in previous studies to characterize the visual attention of children exploring social objects [1]. All data were analyzed and trends plotted, revealing that the most significant scanpaths were those strongly trending toward social stimuli. [Figure 1, Supp]. Our method provides a useful tool for eye tracking researchers working with multiple participants observing identical stimuli.


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.