multimatch-gaze: The MultiMatch algorithm for gaze path comparison in Python

Wagner, Adina; Halchenko, Yaroslav O.; Hanke, Michael

doi:10.21105/joss.01525

Cited by 11 publications

(5 citation statements)

References 8 publications

(7 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(2010) , available at https://seis.bristol.ac.uk/~psidg/ScanMatch/ . For MultiMatch, we use the python reimplementation of the original MATLAB toolbox by Wagner et al. (2019) , available at https://github.com/adswa/multimatch_gaze .…”

Section: Scanpath Similarity Scores On Mit1003mentioning

confidence: 99%

DeepGaze III: Modeling free-viewing human scanpaths with deep learning

2022

View full text Add to dashboard Cite

Section: Scanpath Similarity Scores On Mit1003mentioning

confidence: 99%

DeepGaze III: Modeling free-viewing human scanpaths with deep learning

2022

View full text Add to dashboard Cite

“…In addition, after each training trial within the VR environment, participants received automated feedback in form of a “similarity value”, a quantitative measurement of how closely their real gaze behavior matched the suggested gaze pattern. This quantitative measure of similarity between gaze behavior and suggested gaze pattern is evaluated at run-time using a Multimatch algorithm [ 32 ], and is described in detail in Appendix A in S1 Appendix .…”

Section: Methodsmentioning

confidence: 99%

Influence of open-source virtual-reality based gaze training on navigation performance in Retinitis pigmentosa patients in a crossover randomized controlled trial

Neugebauer,

Sipatchin,

Stingl

et al. 2024

PLoS ONE

View full text Add to dashboard Cite

Methods A group of RP patients (n = 8, aged 20-60) participated in a study consisting of two 4-week-phases, both carried out by the same patient group in randomized order: In the ‘training phase’, participants carried out a Virtual-Reality gaze training for 30 minutes per day; In the ‘control phase’, no training occurred. Before and after each phase, participants were tasked to move through a randomized real-world obstacle course. Navigation performance in the obstacle course as well as eye-tracking data during the trials were evaluated. The study is registered at the German Clinical Trials Register (DRKS) with the ID DRKS00032628. Results On average, the time required to move through the obstacle course decreased by 17.0% after the training phase, the number of collisions decreased by 50.0%. Both effects are significantly higher than those found in the control phase (p < 0.001 for required time, p = 0.0165 for number of collisions), with the required time decreasing by 5.9% and number of collisions decreasing by 10.4% after the control phase. The average visual area observed by participants increases by 4.41% after training, however the effect is not found to be significantly higher than in the control phase (p = 0.394). Conclusion The performance increase over the training phase significantly surpasses the natural learning effect found in the control phase, suggesting that Virtual-Reality based gaze training can have a positive effect on real-world navigation tasks for patients with RP. The training is available as work-in-progress open-source software.

show abstract

“…In addition, after each training trial within the VR environment, participants received automated feedback in form of a ”similarity value”, a quantitative measurement of how closely their real gaze behavior matched the suggested Scanpath. This quantitative measure of similarity between gaze behavior and suggested Scanpath is evaluated at run-time using a Multimatch algorithm [26], and is described in detail in Appendix A (S1 Appendix). The process behind choosing the pattern for the Scanpath, as well as the decision to make the execution of the Scanpath a voluntary option rather than a necessary part of training, is described in the discussion.…”

Section: Methodsmentioning

confidence: 99%

“…Scanpath is evaluated at run-time using a Multimatch algorithm [26], and is described in detail in Appendix A (S1 Appendix). The process behind choosing the pattern for the Scanpath, as well as the decision to make the execution of the Scanpath a voluntary option rather than a necessary part of training, is described in the discussion.…”

Section: Scanpath Behaviormentioning

confidence: 99%

Influence of open-source virtual-reality based gaze training on navigation performance in Retinitis pigmentosa patients in a crossover randomized controlled trial

Neugebauer,

Sipatchin,

Stingl

et al. 2023

Preprint

View full text Add to dashboard Cite

MotivationPeople living with tunnel vision caused by Retinitis pigmentosa (RP) oftentimes face challenges navigating and avoiding obstacles due to a severely decreased Visual Field. Previous studies have shown that gaze training – guided tasks to encourage specific gaze behavior – help improve visual performance in patients with limited Visual Field. We propose a Virtual-Reality based, at-home gaze training – to the best of our knowledge the first of its kind – to increase effective eye movements and navigation performance of RP patients in a real-world obstacle course.MethodsA group of RP patients (n=8) participated in a study consisting of two 4-week-phases, both carried out by the same patient group in randomized order: In the ‘training phase’, participants carried out a Virtual-Reality gaze training for 30 minutes per day; In the ‘control phase’, no training occurred. Before and after each phase, participants were tasked to move through a randomized real-world obstacle course. Navigation performance in the obstacle course as well as eye-tracking data during the trials were evaluated. The study is registered at the German Clinical Trials Register (DRKS) with the ID DRKS00032628.ResultsOn average, the time required to move through the obstacle course decreased by 17.0% after the training phase (p <0.001), the number of collisions decreased by 50.0% (p <0.001) and the average visual area observed by participants increased by 4.41% (p <0.001). In comparison, after the control phase, the time required to move through the obstacle course was found to have decreased by 5.9% (p= 0.003), collisions decreased by 10.4% (p= 0.609) and the observed visual area increased by 2.06% (p= 0.0835).ConclusionThe performance increase over the training phase significantly surpasses the natural learning effect found in the control phase, suggesting that Virtual-Reality based gaze training can improve real-world navigation performance and effective gaze movements of patients with RP. The training is available as work-in-progress open-source software.

show abstract

multimatch-gaze: The MultiMatch algorithm for gaze path comparison in Python

Cited by 11 publications

References 8 publications

DeepGaze III: Modeling free-viewing human scanpaths with deep learning

DeepGaze III: Modeling free-viewing human scanpaths with deep learning

Influence of open-source virtual-reality based gaze training on navigation performance in Retinitis pigmentosa patients in a crossover randomized controlled trial

Influence of open-source virtual-reality based gaze training on navigation performance in Retinitis pigmentosa patients in a crossover randomized controlled trial

Contact Info

Product

Resources

About