Participants
A total of 61 individuals aged 14–21, including 31 ASD and 30 TD individuals, participated in this study. All participants scored a full intelligence quotient (IQ) of ≥ 80 measured by either the short form of Hong Kong Wechsler Intelligence Scale for Children – fourth edition [WISC-IV-HK:SF; (33)] for participants aged below 15 years and 11 months, or the short form of Wechsler Adult intelligence Scale – fourth edition [WAIS-IV-HK:SF; (34)] for those who were aged 16 or above. Given previous research has shown that ASD individuals exhibit sex-dependent differences in non-social cognitive domains involving executive function (35), we only included males in our sample. Furthermore, as prior studies have also shown that handedness influenced the brain network organization (36), we only included right-handed individuals. The diagnosis of ASD of participants was confirmed by Autism Diagnostic Interview – Revised [ADI-R; (37)]. The social functioning of TD individuals was screened by the second edition of the Social Responsiveness Scale [SRS-2; (38)], with all included TD individuals obtained a T-score equal to or below 59 out of a maximum T-score of 90, indicating that they had normal daily social functioning.
Experimental Procedure
All participants underwent three parts of assessment, including IQ assessment, neuropsychological test and EEG measurement. The sequence of assessments was counterbalanced across subjects to minimize order effects. Parents of all participants from the ASD and TD groups were asked to complete SRS-2. Additionally, parents of participants with ASD were required to complete ADI-R structured interview. The WISC/WAIS-IV-HK:SF and ADI-R structured interviews were administered by an educational psychologist (F.P.) and a child psychiatrist (C.Y.) who were blinded to the study hypothesis respectively. The neuropsychological test and EEG measurement were conducted by the first and second authors.
Neuropsychological Measurement
CANTAB Multitasking Task (MTT) was adopted in this study. The MTT (Fig. 1) measures multiple EF abilities including interference control, set-shifting and set-maintenance. In each test trial, an arrow is randomly displayed on a 10.2” iPad screen, which appears on either the left or right side of the screen with its direction pointing to either the left or right side. Participants are instructed to indicate the direction or the side of the arrow according to the rules displayed on the top of the screen. MTT consists of three blocks. The rules (i.e. direction/side) are unchanged (i.e. non-switching blocks) in the first two blocks, and changes from trial to trial in a randomized order in the third block (i.e. switching block). Some trials display congruent stimuli (i.e. congruent trials; e.g., arrow on the right side pointing to the right) whereas other trials display incongruent stimuli (i.e. incongruent trials; e.g., arrow on the right side of the screen pointing to the left). A total of eight parameters are measured, with response latencies and error rates during congruent trials indicating participants’ performance served as controlled condition, incongruent trials indicating his performance in interference control, switching block indicating his performance in set-shifting and non-switching block indicating his performance in set maintenance.
Electroencephalography (Eeg) Measurement
Neurophysiological states of participants were quantified using EEG. EEG data were obtained using a 24-electrode wireless EEG system [DSI-24; (39)], with electrodes positioned according to the 10–20 system (40). Electrode impedances were maintained at ≤ 10 kΩ. Participants were instructed to sit still and focus their attention on an object (i.e. a car) displayed on a computer monitor with their eyes open for five minutes (41), with the EEG signals sampled at 300 samples per second (yielding 90000 samples for each participant), which were referenced to linked earlobes (42, 43). Given numerous evidence that associates EEG theta band (4–7.5 Hz) signals with attention and executive functions [see (44) for a review], we focused our EEG analysis on theta band in this study.
Eeg Data Preprocessing
The EEG data were preprocessed with EEGLAB Toolbox (45) using MATLAB® R2019a (Natick, Massachusetts, The MathWorks Inc). Data from the 19 electrode positions (Fp1, Fp2, F3, F4, F7, F8, Fz, T3, T4, T5, T6, C3, C4, Cz, P3, P4, Pz, O1, and O2) were used for analysis. Finite Impulse Response (FIR) filtering was conducted with lower edge of the frequency pass band set at 1Hz and higher edge of the frequency pass band set at 30Hz. To filter artifacts, visual inspection of EEG data was first performed to remove artifacts induced by excessive body movements, followed by conducting Independent Component Analysis (ICA) to further remove other artifacts. Components exceeding a probability of 0.9 to be an artifact of any kind (i.e. eye blink, muscle, heart, channel noise, others) detected by ICA were rejected.
Calculation Of Eeg Parameters
fE/I calculation. Notably, the calculation of fE/I is based on long-range temporal correlations (LRTC) represented by detrended fluctuation analysis (DFA) exponent, an index that reflect critical oscillation dynamics (46, 47). Given DFA exponent that lies between 0.5 to 1 corresponds to LRTC of a power law form (48, 49) and DFA exponent equal to or below 0.6 showed nonsignificant LRTC (19), only participants with a DFA exponent above 0.6 and below 1 were included in the calculation of fE/I. Individual DFA exponent data were reported in Table S1. The pipeline for the calculation of fE/I is adopted from Bruining, Hardstone (19) and is illustrated below. To minimize the effect of sampling frequency limited by the EEG recording equipment (50, 51), 4th-order causal bandpass filter was applied to extract data from theta frequency band (4-7.5 Hz). Then, Hilbert Transform was applied to extract the amplitude envelope. The cumulative sum of the amplitude envelope formed the signal profile of each individual. Each window of signal profile was then divided by original amplitude, yielding amplitude normalized signal profile windows. The window size of 5 seconds with 80% overlap was used for the calculation of fE/I. The normalized signal profile windows were detrended, which allowed the calculation of the fluctuation for each window. Finally, the correlation between amplitude and fluctuation was calculated for each EEG channel to yield fE/I.
Oscillation power calculation. Given oscillation power is the most common way to characterize resting-state EEG and some previous studies have revealed band-specific power abnormality in ASD (52), we supplemented the fE/I analysis with theta oscillation power analysis. Absolute power was calculated using the Welch’s method with a 2-second Hamming window and a frequency resolution of 0.01Hz. Relative power was computed by dividing the absolute power in the theta band by the broadband power in the range of 1-30Hz.
To increase the signal-to-noise ratio of EEG data, fE/I and absolute/relative power values for individual channels were averaged within different regions-of-interest (ROIs), namely global (average of fE/I from all 19 scalp channels), left (F3, Fp1, F7, C3, P3, T3, T5, O1), midline (Fz, Cz, Pz), right (F4, Fp2, F8, C4, P4, T6, T4, O2).
Data analysis
To compare the baseline demographic characteristics (i.e. age, IQ and SRS-2 total score) between ASD and TD groups, independent sample t-tests were performed. To compare the performance between ASD and TD in baseline condition, interference control, set-maintenance and set-shifting, error terms and reaction times during congruent, incongruent trials, as well as switching and non-switching blocks, were analyzed with independent t-tests (or Mann-Whitney U tests when normality of data was not achieved) with Bonferroni corrections. With alpha-level kept at 0.05, a significant between-group difference in these neuropsychological parameters was indicated by p < 0.05/8 = 0.00625. To compare the theta functional fE/I, absolute and relative power between ASD and TD groups, independent t-tests were performed. With alpha-level kept at 0.05, a significant between-group difference for 4 ROIs was indicated by p < 0.05/4 = 0.0125 for each of the parameters (i.e. fE/I, absolute power, relative power). To explore the relationship between the neurophysiological parameters, clinical characteristics and basic cognitive and EF functioning, Pearson’s correlations (two-tailed) were separately conducted for TD and ASD groups for parameters with significant between-group differences, with the false discovery rate (FDR) maintained at alpha-level of 0.05 using the Benjamini–Hochberg procedure (53).