Reliable Single-Trial Detection of Saccade-Related Lambda Responses with Independent Component Analysis

In natural, free-viewing settings, visual perception is driven by a series of saccades and fixations. Perceptual mechanisms are typically studied through averaged fixation-related potentials generated from simultaneous eye-tracking and EEG recordings. Lambda responses following fixation onsets signal the arrival of new visual input to the primary visual cortex. In our study, we investigate the use and preprocessing parameter dependence of independent component analysis (ICA) in separating the lambda response from other neural sources. In our experiment, 10 subjects (2 males and 8 females) viewed 80 art paintings in natural, free-viewing settings, during which EEG data were recorded. Our results show that unique lambda response components can be detected reliably and individual lambda waves can be extracted in a single-trial manner, without signal averaging. ICA decomposition is most sensitive to high-pass filtering producing best results with a minimum 1 Hz filtering. We also propose a method that automatically and accurately identifies the lambda component among other independent components for further lambda peak detection. These individual lambda waves can then be used to study saccade-related modulation effects without losing temporal and spatial resolution. The novelty of our method is the automatic detection of lambda components and extraction lambda waves, which is a new approach in saccade/fixation and visual perception research under naturalistic viewing conditions.