The experimental study included 21 patients affected by early PD (age = 62.52 ± 9.6) and 22 healthy subjects (age = 59.06 ± 9.7) as control group. PD patients, whose clinical and demographic characteristics are reported in Table 1, were enrolled during routine clinical practice, at the Neurophysiopathology Unit of Bari Polyclinic General Hospital. Inclusion criteria were: diagnosis of idiopathic Parkinson’s disease, Hoehn-Yahr stage < II, age between 40–80 years, MMSE > 24, absence of significant visual deficits. All patients were stable without motor/non-motor fluctuations and dyskinesias.
All participants were consistent right-handers according to the Edinburgh Handedness Inventory. All participants provided written fully informed consent before entering the experimental study. The study was conducted in compliance with the Declaration of Helsinki and approved by the Ethics Committee of the Bari Polyclinic General Hospital. For each group, subjects with less than eight years of schooling, with major psychiatric diseases, diseases of the central or peripheral nervous system, diabetes, thyroid diseases, severe chronic kidney disease, autoimmune and connective system diseases, were considered not eligible for the study, together with cases using substances or drugs with CNS effects, excluding those specifically prescribed for PD (L-dopa, dopa agonists).
We adopted the same stimuli of the study by Craighero et al.7. They consisted of two videos of the same duration (2640 ms), showing the same agent sitting at a desk executing a movement for reaching and grasping an object. The agent was recorded from a third-person perspective. In the “flat object video” the object consisted of a parallelepiped (width: 7 cm; height: 3 cm; length: 3 cm) placed with its longer axis facing the agent. Without lifting the object, the agent reached out and naturally grabbed the parallelepiped with her fingers parallel to her frontal plane. Using a software for video editing, the parallelepiped was changed to a polyhedron (with the same dimension of the parallelepiped, i.e., 7 cm × 3 cm × 3 cm) in the “sharp-tip object video”. In this way, the kinematics of the movement remained the same and the agent’s fingers touched the object precisely at the sharp tips. The instant at which the experimenter’s index finger touched the object was the same for both videos (1880 ms, Frame 47). The two videos were further modified to produce two catch-trials videos stopped before the agent’s hand touched the objects (1520 ms after the beginning of the video, Frame 38). To achieve the same time duration as the experimental-trial videos (2640 ms), the last frame was repeatedly shown. Catch-trial videos were provided merely to make participants constantly pay attention to video content all the time; they were never taken into account while analyzing the results.
A multimodal fNIRS-EEG co-registration system was used to conduct the study, as detailed in the following paragraphs. The participant was seated on a chair in a quiet room, in front of a desk on which there was a display (60 cm from the person) and a keyboard. Before the experiment began, each participant was asked to grab both objects shown in the video with the same grip used by the agent. This request was intended to demonstrate to the participants that the heavy weight of the object (240 g), and the presence of sharp tips right at the point of contact with the fingers made it impossible to grasp the sharp-tip object with the grip shown in the video. On the contrary, although the weight was the same, that type of grip was suitable for grasping the flat object. Participants were submitted to two experimental sessions: time-to-contact detection session, and observation-only session. Each experimental session consisted of 42 randomly presented trials: 30 experimental trials (15 flat object videos, and 15 sharp-tip object videos) and 12 catch trials (6 flat object catch videos and 6 sharp-tip object catch videos).
At the beginning of each experimental session, the participant stared at a fixed cross on the black screen for 120 s, to record 20 s baseline and 100 s resting state. They were informed about the type of session which was starting at the beginning of the resting-state recording, and warned again 5 s before the start of the videos. In fact, we intended to use the resting-state to detect, for both fNIRS and EEG, the modality of preparation to action observation. A black screen was shown between videos for 15 s. At the end of each experimental session, each participant was allowed 5 min of rest.
Time-to-contact detection session
The participant’s left arm was relaxed on the desk. They were told to watch the videos and use their right index finger to tap the space bar on the keyboard when the agent touched the target object (experimental trials); however, they were not allowed to tap the space bar when the agent’s hand stopped before touching the target object (catch trials). The participant’s response to catch trials was counted and treated as an error. The subject was excluded from the analysis if the number of errors was 6 or higher. A small percentage of incorrect responses made sure that the task was executed considering the time of the touch and no further clues.
Both of the participant’s arms were relaxed on the desk. Participants were instructed to watch the videos carefully. Six times, at random, the following question appeared on the screen: “What object did you just see?”. Participants’ responses were noted and verified by a researcher. If the participant made a number of errors equal to or greater than 3, they were excluded from the study sample. The error limit ensured that only participants with high levels of attention to the videos were considered.
We used a cap adapted to co-registration consisting of 20 fNIRS channels and 62 active Ag–AgCl surface electrodes (Fig. 8).
For this study, a continuous wave NIRS system (NIRSport 8 × 8, Nirx Medical Technologies LLC, Berlin, Germany) that captures brain oxygenation measurements was adopted. The fNIRS device is considered a promising tool for studying action observation mechanisms25. The device is a multi-channel system, easy to wear, and contains LED sources and photosensitive detectors (sensitivity: > 1 pW, dynamic range: > 50 dB). The data was recorded using NIRStar 14.2 software (Version 14, Revision 2, Release Build, 2016-04-15 NIRx Medizintechnik GmbH, Berlin, Germany; www.nirx.net). The experimental design involved the use of eight sources and eight detectors (commonly known as optodes). The NIR light source used in the system emits two wavelengths of 760 and 850 nm. The light absorption by brain tissues is wavelength-dependent and permits to estimate of different values for the oxyhemoglobin (ΔHbO2) and deoxyhemoglobin (ΔHb) respective concentrations during experimental sessions. A wavelength-dependent differential path length factor (DPF) is included in the modified Beer-Lambert law (MBLL) to simply determine the relative variations of the concentration of ΔHbO2 and ΔHb.
The optodes were positioned on selected portions of the scalp with the intent of evaluating the activity of a set of cerebral cortical regions relevant to the designed experiment; specifically, they were positioned over the primary and supplementary motor cortex (as shown in Fig. 8). As the previous evidence suggests, the ideal distance between sources and detectors was 30 mm for acquiring a good optical signal. The oxyhemoglobin and deoxyhemoglobin (ΔHbO2 and ΔHb) data were collected with a sampling rate of 7.81 Hz. However, in this study, we evaluated the ΔHbO2 levels only (see paragraph below). A sensor calibration procedure was carried out before every measurement. This digital practice allows ascertaining the adequate signal amplification for each optodes combination.
fNIRS signal processing
fNIRS signal processing was made using nirsLAB MATLAB-based software (nirsLAB, version 2017.06, NIRx Medical Technologies, Glen Head, NY, USA). Specifically, the researchers performed a raw signal cleaning process with the following functions: removing discontinuities from the signal, removing peak motion artifacts, baseline correction, and hemoglobin molar extinction coefficients.
The analysis was carried out on 8-s-long epochs, where each epoch began at the start of each video. For digital filtering, a 6th-order Butterworth Low Pass Filter with cut-off frequencies of 0.06–0.2 Hz was applied to raw data to eliminate the respiratory and cardiac frequencies from the signal in each recording channel. The oxyhemoglobin levels changes were computed, as a reliable measure of cortical metabolic status26. To calculate the molar extinction coefficients of hemoglobin, the researchers adopted the spectrum published by W.B. Gratzer (Med. Res. Council Labs, Holly Hill, London) and N. Kollias (Wellman Laboratories, Harvard Medical School, Boston, MA, USA). Then, the optical intensity data were converted into ∆HbO2 (in mmol/liter) concentration by the modified Beer-Lambert law.
Before the application of the modified Beer-Lambert law, for each participant, a baseline corresponding to the first 20 s of each recording was subtracted. Block-average fNIRS responses were then calculated, after subtracting baseline averages of each block, i.e., 5 s before stimulus onset.
The EEG signal was acquired using the Micromed Brain Quick equipment at a sampling rate of 256 Hz using 61 electrodes positioned according to the 10–10 international system. To acquire also the electrooculogram (EOG), two electrodes were placed on the right and left eyebrows, respectively. The reference electrode was positioned on the nasion, and the ground electrode on Fpz. A 0.1–70 Hz band-pass filter with a 50 Hz digital filter was applied during the EEG recording. The EEG was recorded during the entire experimental procedure.
EEG signal processing
The EEG data preprocessing was performed by adopting EEGLAB 14.1.1, a MATLAB-based software27. The researchers used a semi-automatic method based on visual detection and channel statistics to locate and remove the faulty recording channels. All channels with distributions far from the Gaussian one were excluded from the analyses. Ocular artifacts recorded by the EOG channels were removed by means of the ICA algorithm included in the EEGLAB tool. Next, all the EEG files were processed using Letswave 7 tool (https://letswave.cn). EEG has been re-referenced to 0 value and pre-filtered with a band-pass filter in the range[1–24] Hz.
To evaluate the not phase-locked synchronization/desynchronization of alpha mu and beta mu, the researchers used a time–frequency analysis based on Continuous Wavelet Transform (CWT)28, with a baseline correction computed on the 20 s preceding the resting-state. The absolute power of the alpha (7–12 Hz) and beta (13–30 Hz) bands were considered for the single experimental conditions.
In order to detect the preparation to action observation, epochs lasting 5 s that preceded both the start of the observation-only session and the time-to-contact detection session were considered. To evaluate the alpha mu changes, with respect to the baseline, related to the vision of the flat and sharp tip objects, we computed the CWT in a time window from 2 s preceding the object grasping to 1 s following it, so the EEG was recorded simultaneously to the movement of the arm in the video.
The considered dependent variable was the time lag between the instant at which the agent’s index finger touched the object (Instant of Touch), i.e., 1880 ms from the beginning of each video, and the participant’s key pressing (Response). For each participant, for each trial, we calculated the time lag as Instant of Touch-Response. The time lag was submitted to a two-way ANOVA with Object (flat object vs sharp-tip object) as the within-subject variable and Group (PD patients vs Controls) as the between-subjects variable.
Before starting each recording, the participant’s age-dependent Differential Path-length Factor (DPF) was entered into the signal acquisition software. To study changes in cerebral hemodynamic activity during tasks, the mean of the change in ∆HbO2 was calculated for each experimental condition. According to our previous study11, the analysis window chosen for each event was 8 s. The GLM method was applied to investigate the activation in brain regions of interest. Therefore, for each experimental session, the hemodynamic response function (HRF) was adopted to model the fNIRS signal. SPM-1 within-subject analysis allowed for estimating the activation (beta values) in each fNIRS channel with respect to the baseline. A two-way ANOVA was also applied to evaluate the oxyhemoglobin levels on the averaged values of left and right fNIRS channels, using groups and sessions as factors in the resting state, and groups and objects as factors in the observation-only and time-to-contact sessions.
For behavioral data and Oxyhemoglobin levels on single channels, we used the IBM Statistics Package for the Social Sciences (SPSS), Version 28.0 (IBM Corp., Armonk, NY, USA). For all analyses, the significance level was set at 0.05.
For topographical analysis and generation of Statistical Probability Maps, we used Matlab Letswave 7 tool, applying the Student’s t-test for paired data to compare the absolute power of alpha power in single groups. The two-way ANOVA with conditions and groups as factors was also applied, to establish differences of alpha mu behaviour in resting-state preceding observation and time-to-contact detection sessions and during these sessions between flat vs sharp tip object grasping conditions. To overcome the multiple comparisons problem, we performed the statistical analyses performing a nonparametric cluster-based permutation approach29. The calculation of the cluster-based statistics consists in grouping together neighboring t-values obtained for (frequency, time)-samples into clusters and summing the statistical values within each cluster. For inclusion in a cluster, only statistical values higher than the cluster-forming threshold, which was set to 0.05, are considered. Then, the significance probability is calculated with a Monte Carlo approximation based on the number of permutations. As a rule of thumb proposed in previous works30,31, this number should be no less than 1000. Thus, to perform feasible computations, we set it to 2000. For representation purposes, the Statistical Probability Maps show the significant results obtained after permutations in the range 0.001–0.01.