Application of Time-Scale Decomposition of Entropy for Eye Movement Analysis.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Data Science

Background:

• Eye movement analysis is crucial for understanding neurological function and diagnosing disorders.
• Traditional methods may not fully capture the complex dynamics of eye movement signals.
• Nonlinear time series analysis offers advanced tools for characterizing complex biological signals.

Purpose of the Study:

• To investigate the utility of nonlinear time series analysis measures for characterizing eye movement signals.
• To apply these measures in an eye movement event detection and classification task.
• To enhance the accuracy of detecting saccadic latency and saccades.

Main Methods:

• Employed nonlinear time series analysis techniques, including approximate entropy, fuzzy entropy, and Largest Lyapunov Exponent.
• Defined multilevel maps (MMs) as time-scale decompositions of these measures.
• Utilized k-Nearest Neighbors (kNN) classification with feature vectors derived from MM segments.

Main Results:

• Feature vectors incorporating elements from multilevel maps and nonlinear measures improved classification accuracy.
• Significant improvements were observed in the accuracy of saccadic latency and saccade detection.
• The proposed method demonstrated superior performance compared to previous studies on eye movement dynamics.

Conclusions:

• Nonlinear time series analysis, particularly using multilevel maps, is effective for eye movement signal characterization.
• These advanced analytical methods enhance the precision of eye movement event detection.
• The findings suggest potential for improved diagnostic tools in ophthalmology and neurology.