2.3 EEG data acquisition and preprocessing
EEG was continuously recorded from a 64-electrode EEG cap (Neuroscan Quikcap) via a Neuroscan SynAmps2 amplifier using Scan 4.5 software (Compumedics Neuroscan, USA; sampling rate: 1 kHz, DC-200 Hz) while subjects performed the SORT task. The reference electrode was placed between Cz and CPz at the midline. A bipolar vertical electrooculogram (VEOG) was recorded for the left eye. EEG data from electrodes with impedance exceeding 10–20 kΩ were discarded from further processing and most impedances were less than 5–10 kΩ. Poorly functioning electrodes were excluded manually by visual inspection of the raw data. Data from fewer than 5% of electrodes were rejected; the number of rejected electrodes did not differ significantly across groups (YA vs HA,t (49) = 0.15, p = 0.881; HA vs MCI, t (40) = 0.44,p = 0.663). An algorithm computing the average based on spherical splines fitted to the data was then applied to interpolate EEG data to the sites of the bad electrodes identified (Ferree et al., 2009).
The continuous EEG data were high-pass filtered at 1 Hz and subsequently low-pass filtered at 40 Hz using a finite impulse response filter. Independent component analysis (ICA) was then applied to the filtered continuous EEG data to identify artifacts (muscle, eye, and heart) using the EEGLAB toolbox (Delorme and Makeig, 2004) running under MATLAB 2021 (MathWorks, Natick, MA, USA). Components with >70% probability of representing an artifact were automatically removed (ICLabel, Pion-Tonachini et al., 2019). Subsequently, ICA components of each individual’s data were visually examined and artifacts not identified previously by the algorithm were removed manually to complete data cleaning. After this step, EEG data were segmented for retrieval and non-retrieval conditions into multiple EEG epochs (-1000 to 3000 ms, time-locked to the stimulus onset). Epochs having a peak amplitude of ±75 µV between -500ms and 2000 ms post stimulus onset (highly associated with artifacts) were rejected and epochs with extreme values (based on standard deviation > 5) were excluded by rejection algorithms in EEGLAB using joint probability and kurtosis of activity. The average (accepted trials / all correct trials) epochs retained for retrieval and non-retrieval conditions respectively were 88.3 ± 5.8 % (42.3 ± 4.1 trials) and 89 ± 6.6 % (42.7 ± 5.2 trials) in the YA group, 91 ± 5.3 %(44.7 ± 4.2 trials) and 91.9 ± 4.1 % (44.8 ± 4.4 trials) in the HA group, and 88.6 ± 5.8% (39.3 ± 4 trials) and 88.3 7.5 ±% (40.3 ± 7.6 trials) in the MCI group.