Effective correction of extreme capacitive artifacts in TMS-EEG via windowed detrending

Structured AbstractO_ST_ABSBackgroundC_ST_ABSTranscranial Magnetic Stimulation (TMS) combined with electroencephalography (EEG) is a tool for investigating non- invasively cortical excitability and connectivity. However, this technique is susceptible to artifacts that can obscure genuine neural activity, making signal correction a critical component of data processing. Among these, capacitive artifacts pose a significant challenge, as they hinder the accurate interpretation of TMS-evoked potentials (TEPs). These artifacts can often be mitigated online through optimized recording settings and experimental procedures, but in the most challenging cases they must be removed offline to ensure accurate reconstruction of TEPs. MethodsTo correct capacitive artifacts, we tested detrending methods based on both windowed and non-windowed approaches. Non-windowed methods operate on the full EEG epoch, while windowed approaches segment the signal into rise and decay phases of the post-stimulus signal, which correspond to the electrode-related charging and discharging effects triggered by the stimulation. We applied these methods on several datasets from two centers acquired with different hardware settings. We controlled capacitive artifacts and benchmarked detrending algorithm performance against a cleaning strategy based on Independent Component Analysis (ICA). ResultsCapacitive artifacts varied from mild to extreme, depending on the amplitude and characteristic time constants of the capacitive model. ICA was effective in mild cases but failed to adequately correct moderate to severe artifacts. In contrast, model-based detrending approaches, particularly windowed methods, proved more effective in managing extreme artifacts, as segmented fitting facilitates parameter identification. Among these, windowed polynomial detrending showed a slight advantage due to its adaptability to complex artifact shapes. ConclusionsWhile the most effective way to minimize TMS-related capacitive artifacts is to use optimal TMS-compatible hardware together with rigorous online prevention and monitoring during acquisition, reliable offline correction methods remain essential when ideal recording conditions cannot be guaranteed. In this context, windowed detrending provides a robust offline strategy that improves the reliability of TEP interpretation, particularly in settings where hardware optimization is not feasible. HighlightsO_LICapacitive artifacts in TMS-EEG recordings vary in severity and can compromise the interpretation of TMS- evoked potentials (TEPs) C_LIO_LIWindowed detrending outperforms ICA and non-windowed models in correcting extreme capacitive artifacts and restoring the physiological 1/f spectral profile. C_LIO_LIAdvanced signal correction methods enhance the reliability of TMS-EEG recordings, particularly when artifacts are source-related or when affected channels cannot be excluded. C_LI