Saccade detection using a particle filter.

Area of Science:

• Ophthalmology
• Neuroscience
• Biomedical Engineering

Background:

• Healthy subjects use catch-up saccades to compensate for imperfect tracking gain during smooth pursuit.
• Evaluating saccade and pursuit kinematics necessitates separating these components from eye movement data.
• Analyzing patient eye movements is challenging due to potential drifts between saccades and a wide range of saccade amplitudes (microsaccades to 40°).

Purpose of the Study:

• To introduce a novel algorithm for robust saccade detection.
• To address limitations of traditional methods in handling smooth pursuit and eye drifts.
• To provide a reliable tool for analyzing complex eye movement data.

Main Methods:

• A particle filter-based algorithm was developed for saccade detection.
• The method suppresses baseline velocity from smooth pursuit or eye drifts, allowing a constant detection threshold.
• The algorithm accommodates a broad spectrum of saccade amplitudes.

Main Results:

• The new method demonstrated effectiveness across five paradigms: microsaccades, saccades with drift, linear, non-linear target motion, and free viewing.
• Sensitivity to signal noise was systematically analyzed.
• Validation confirmed the algorithm's ability to detect saccades in challenging scenarios.

Conclusions:

• The proposed particle filter algorithm reliably detects saccades during smooth pursuit and/or eye drifts.
• It offers an improvement over traditional threshold-based methods for unfiltered eye movement recordings.
• This facilitates more accurate analysis of eye movements in both healthy and patient populations.