Computational Methods for Continuous Eye-Tracking Perimetry Based on Spatio-Temporal Integration and a Deep Recurrent Neural Network.

Alessandro Grillini,Frans W. Cornelissen,Giorgia Demaria,Remco J. Renken,Alex Hernández-García

doi:10.3389/fnins.2021.650540

Abstract

The measurement of retinal sensitivity at different visual field locations–perimetry–is a fundamental procedure in ophthalmology. The most common technique for this scope, the Standard Automated Perimetry, suffers from several issues that make it less suitable to test specific clinical populations: it can be tedious, it requires motor manual feedback, and requires from the patient high levels of compliance. Previous studies attempted to create user-friendlier alternatives to Standard Automated Perimetry by employing eye movements reaction times as a substitute for manual responses while keeping the fixed-grid stimuli presentation typical of Standard Automated Perimetry. This approach, however, does not take advantage of the high spatial and temporal resolution enabled by the use of eye-tracking. In this study, we introduce a novel eye-tracking method to perform high-resolution perimetry. This method is based on the continuous gaze-tracking of a stimulus moving along a pseudo-random walk interleaved with saccadic jumps. We then propose two computational methods to obtain visual field maps from the continuous gaze-tracking data: the first is based on the spatio-temporal integration of ocular positional deviations using the threshold free cluster enhancement (TFCE) algorithm; the second is based on using simulated visual field defects to train a deep recurrent neural network (RNN). These two methods have complementary qualities: the TFCE is neurophysiologically plausible and its output significantly correlates with Standard Automated Perimetry performed with the Humphrey Field Analyzer, while the RNN accuracy significantly outperformed the TFCE in reconstructing the simulated scotomas but did not translate as well to the clinical data from glaucoma patients. While both of these methods require further optimization, they show the potential for a more patient-friendly alternative to Standard Automated Perimetry.

Highlights

The assessment of the quality of the visual field is a staple of ophthalmologic evaluation
Evaluation on Simulated Visual Field Defects After reconstructing the visual field maps, we evaluated the performance of the threshold-free cluster enhancement (TFCE) method and recurrent neural network (RNN) “time-point classifier” by computing a 2D Spearman rank correlation between the ground-truth maps and the reconstructed maps
The results of the optimization of the parameter λn based on the maximum average accuracy of each condition for each fold are shown in Figure 6, while Figure 7 shows the effect that adjusting λn has on the visual field map reconstruction of a participant across all conditions

Summary

Introduction

The assessment of the quality of the visual field ( called perimetry) is a staple of ophthalmologic evaluation. Patient performance is affected by learning (Schultz, 1990; Wild et al, 2006) and fatigue (Johnson et al, 1988), as well as the expertise of the operator (Montolio et al, 2012). Together, these constraints limit the effectiveness of SAP, in clinical and rehabilitation contexts such as when dealing with children (Walters et al, 2012), the elderly, and/or cognitively impaired patients (Diniz-Filho et al, 2017; Gangeddula et al, 2017)

Methods

Results

Discussion

Conclusion