Decomposition frequency optimization in wavelet-based template matching algorithms to manage P300 latency jitter

• Published in: 2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) Vol. 2024; pp. 1 - 4
• Main Authors: Quattrociocchi, Ilaria, Caracci, Valentina, Riccio, Angela, Galiotta, Valentina, D'Ippolito, Mariagrazia, Cincotti, Febo, Toppi, Jlenia, Astolfi, Laura
• Format: Conference Proceeding Journal Article
• Language: English
• Published: United States IEEE 01.07.2024
• Subjects: Adaptive Wavelet Filtering; Algorithms; EEG; Electroencephalography; Electroencephalography - methods; Event Related Potentials; Event-Related Potentials, P300 - physiology; Frequency estimation; Humans; Jitter; Latency Jitter; Optimization; P300; Physiology; Signal Processing, Computer-Assisted; Signal resolution; Signal to noise ratio; Time-domain analysis; Time-frequency analysis; Wavelet Analysis; Wavelet domain
• ISSN: 2694-0604
• DOI: 10.1109/EMBC53108.2024.10782005

Event-Related Potentials (ERPs) are modifications of the brain activity in response to external sensory stimulation. P300 is a positive ERP component that occurs 300 ms after the presentation of a rare stimulus and indicates conscious perception of an unexpected change in sensory stimulation. Since the amplitude of ERPs is comparable to that of background electroencephalography, the most widespread approach to render ERPs correctly detectable is time-domain averaging. ERPs are physiologically affected by variations in latency, referred to as latency jitter. The most promising techniques to address latency jitter are cross-correlation template matching algorithms, which estimate single-trial latencies and resynchronize them according to a template that best describes the ERP component. Adaptive Wavelet Filtering (AWF) is an algorithm applied in the wavelet domain to provide the optimal compromise of time-frequency resolution in detecting ERPs. The choice of an appropriate decomposition frequency is crucial in the application of AWF, as it is essential for identifying the correct template and estimating single-trial latencies. In this study, we investigated the influence of frequency on the performance of AWF to optimize it. The algorithm was tested on simulated data with varying latency shifts and signal-to-noise ratios as well as for data recorded from 11 healthy subjects during an auditory oddball paradigm. Results on both simulated and real EEG data showed how frequency decomposition of 3 Hz is the most appropriate for P300 waveform detection.