Brain responses vary in duration-modeling strategies and challenges

Area of Science:

• Neuroscience
• Cognitive Science
• Data Analysis

Background:

• Event-related brain responses are typically analyzed assuming fixed event durations.
• Varying event durations (e.g., reaction times, stimulus length) and subsequent signal overlap are common in neuroimaging.
• Existing methods for signal overlap correction may be insufficient when event durations differ.

Purpose of the Study:

• To investigate the impact of unmodeled event durations on brain response analysis.
• To propose and evaluate methods for explicitly accounting for event durations in neuroimaging data.
• To assess the compatibility of duration modeling with existing overlap correction techniques.

Main Methods:

• Simulations were used to compare different multiple regression approaches for modeling event durations.
• Non-linear spline regression was employed to capture duration effects.
• The proposed duration modeling was combined with linear overlap correction for electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) data.

Main Results:

• Failure to account for event durations can lead to spurious findings in brain response analysis.
• Non-linear spline regression for duration effects demonstrated superior performance compared to other methods.
• Non-linear event duration modeling is compatible with linear overlap correction, enabling joint application.

Conclusions:

• Explicitly modeling event durations is essential for accurate analysis of brain responses, especially when durations vary across conditions.
• Mass-univariate models incorporating non-linear spline regression for duration and linear overlap correction offer a flexible approach for analyzing overlapping brain signals.
• These findings are applicable to EEG, fMRI, and potentially other overlapping physiological signals like LFPs and pupil dilation.