PELP: accounting for missing data in neural time series by Periodic Estimation of Lost Packets

Objective Implanted electrical stimulators with sensing capabilities have enabled the development of closed-loop neuromodulation therapies capable of responding to patient needs in real-time. Through a combination of rechargeable technologies and wireless data transmission, it is now possible for researchers to acquire extensive neural recordings from human participants in naturalistic settings using these bidirectional devices. However, data losses during wireless transmission hamper processing and the identification of neural signals of interest, driving the need for methodologies to properly estimate the impact of data loss. Approach To accurately reconstruct the timing of data containing losses, we have developed a method called Periodic Estimation of Lost Packets (PELP) to precisely determine the number of samples lost from implanted recordings during active stimulation. PELP leverages a data-driven procedure for determining the period of stimulation and the knowledge that stimulation continues identically during periods where data are missing to accurately account for the number of samples lost. Main results Using simulated stimulation added to collected human EEG data, we show that PELP is robust to a range of stimulation waveforms and noise characteristics. Lastly, we successfully applied PELP to local field potential (LFP) recordings from an implanted, bidirectional device using data recorded in the clinic and the patient’s own home. Significance By effectively accounting for the timing of missing data, PELP enables the analysis of complex, naturalistic neural time series data from bidirectional implanted devices aiding in the development of novel therapeutic approaches. NCT04806516 ( ClinicalTrials.gov ).