Poster #160 - Estimating individual auditory and visual brain responses in 7-month-old infants watching a five-minute cartoon movie

• In Event: 1-005 - Poster Session 01
  In Poster Session: PS 01 Section - Perceptual, Sensory, Motor

Integrative Statement

Electroencephalography (EEG) continues to be the most popular method to investigate cognitive brain mechanisms in young children and infants. Yet, while in adult EEG research, a host of new analysis approaches has been developed over the past decade, most studies in infants still rely on the use of event-related brain potentials (ERPs). ERP-designs and -analyses have the advantage of being well-established and easy-to-use but also have a number of shortcomings. In particular, their computation requires a large number of repetitions of items from the same stimulus-category, responses to which are then averaged and compared between conditions. Especially in young children, this can be a challenging procedure, as their attention to repetitive stimuli needs to be captured for a sufficiently long duration. Furthermore, the ecological validity of ERP-designs is often limited.
We therefore explored a way to investigate infant continuous EEG responses to an ongoing engaging signal (i.e., “neural tracking”) by using multivariate temporal response functions (mTRFs), an approach that has become increasingly popular in adult EEG-research (e.g. Cross et al., 2016, Front Hum Neurosci; O’Sullivan et al., 2017, Front Hum Neurosci). In short, mTRFs rely on the computation of an encoding model that allows for a mapping between the recorded ongoing EEG signal and the stimulus signal.
Fifty-two 7-month-old infants watched a 5-min episode of an age-appropriate cartoon show (“Peppa Pig – The new car”), while we recorded the EEG signal. We extracted two regressors from the video: An auditory regressor containing the amplitude envelope of the soundtrack and a motion regressor based on the mean change per pixel between two consecutive frames. We computed three separate encoding models (auditory only, visual only, and audiovisual). Using a leave-one-out procedure, we calculated a generic model based on training data from n-1 participants and used the obtained model to predict the EEG response of the remaining participant. We compared the generic model to individual models that were cross-validated based on training-data subsets from the available data for each participant.
In the generic model, we observed a clearly defined response function for both, the auditory (Fig 1a) and the motion regressor, characterized by a prominent positive frontocentral deflection peaking around a lag of 400 ms. Importantly, this response profile was visible also on an individual level (Fig 1b). A cluster-based permutation test across channels and time points reveals the statistical significance of the observed response functions compared to zero (Fig 1c). When comparing predictions arising from the generic model to those arising from the individual model, both show a positive correlation with the actual EEG response, which is however higher for the generic model.
In sum, we demonstrate that mTRFs are a feasible way of analyzing continuous EEG responses in infants. This modelling method allows for robust response estimates both across and within participants from only 5 minutes of recorded EEG signal. Our results open way for using more engaging and at the same time more ecologically valid stimulus material when studying cognitive, perceptual, and affective processes in infants and young children.

Authors

• Sarah Jessen, Department of Neurology, University of Luebeck
  Presenting Author

• Lorenz Fiedler, Department of Psychology, University of Luebeck
  Non-Presenting Author

• Thomas Muente, Department of Neurology, University of Lübeck
  Non-Presenting Author

• Jonas Obleser, Department of Psychology, University of Luebeck
  Non-Presenting Author