SmartERD: a pipeline for referencing subjects to a common peak in the analysis of ERD dynamics.

Identifying band-specific features -such as amplitude and latency/duration-from event-related desynchronization (ERD) patterns in MEG/EEG time-frequency representations can enhance our understanding of how the brain processes tasks and influences behavior. However, these features can be affected by potential noise, intrinsic instabilities, and inter-trial variability, which makes it challenging to analyze ERD dynamics accurately when multiple peaks with similar amplitudes emerge in the ERD pattern. To address these issues, we introduce SmartERD, a new pipeline designed to extract dynamic ERD features while considering ERD variability. SmartERD automatically estimates a band-specific pseudo-confidence interval around the absolute ERD peak, inspired by uncertainty propagation principles. It then identifies the first local ERD peak within this interval -representing the first component of the induced oscillatory response- and automatically extracts its features. A key advantage of this approach is that it extracts the latency of a common reference point for all subjects -the first peak response after a trigger- facilitating more consistent comparisons. We validated SmartERD through realistic simulations that mimic different trial numbers. Results showed that SmartERD's estimates are closer to the ground truth across various noise levels compared to standard methods. Additionally, when applied to experimental MEG data, SmartERD demonstrated a better ability to capture individual differences in ERD dynamics, thanks to the selection of a common reference latency. Overall, we propose SmartERD as a valuable tool for extracting meaningful features from oscillatory dynamics, with the potential to improve the analysis of brain-behavior relationships in complex cognitive tasks.