2.1.

Ethics Statement

The study protocol was designed according to the Declaration of Helsinki and reviewed and approved by the Institutional Review Board (registration number 77-21). The study was conducted at the Neurology Department of the Central Hospital “Dr. Ignacio Morones Prieto” in Mexico. The recruiting period was from October 2021 to October 2022. Before testing, all participants provided signed informed consent.

2.2.

Participants

Twenty patients were enrolled in our clinic and were diagnosed according to the criteria of the United Kingdom PD Society Brain Bank.²⁰ All PD patients were on anti-Parkinson drugs. Daily Levodopa equivalent doses were determined from total daily doses for participants using anti-Parkinson medicines using a conversion formula.²¹ In the present analysis, we included patients with a tremor-dominant ($N=11$) or akinetic-rigid subtype ($N=9$).²² A neurologist (FJRR) used the Hoehn and Yahr (HY) scale²³ and the Movement Disorder Society-Sponsored Revision of the Unified Parkinson Disease Rating Scale (MDS-UPDRS, part III) to determine the severity of the disease.³ Freezing of gait severity was assessed with the freezing of gait questionnaire (FOGQ).²⁴ All evaluations detailed in this study were conducted while the subjects were in their “off” medication state ($\sim 12\mathrm{h}$ after the last anti-parkinsonian medication dose was taken). The MDS-UPDRS part III scores were additionally validated by an independent rater.

The control group was composed of 20 age- and sex-matched healthy participants without movement disorders. All participants underwent the Spanish version of the Montreal Cognitive Assessment (MoCA) for global cognitive function assessment.²⁵ Severe cognitive impairment at baseline was defined as a score $<10$ on the MoCA examination, and individuals with this score or below were excluded from this study.

2.3.

fNIRS Data Acquisition

The portable fNIRS system Brite MKII (Artinis Medical Systems BV, the Netherlands) was employed for measuring cortical brain activity in the present study. This system employs 10 dual-wavelength light-emitting diodes (LEDs) centered at 757 and 843 nm, with a sampling frequency of 25 Hz, which is sufficiently high enough to avoid aliasing effects of blood pressure, respiratory, and cardiac frequencies. Participants in the study were outfitted with a black neoprene head cap, with sizes varying between 54 and 60 cm, based on their head circumference, as depicted in Fig. 1(a). The cap was placed on the head of the participant, and the fNIRS optodes were securely held throughout data collection. There were eight detectors and 10 near-infrared sources; the optodes were placed 3 cm apart for the 20 long channels and 1.5 cm apart for the two short channels. The sources and detectors were positioned in the cap holders before the cap was worn by the participant and then manually adjusted to achieve an efficient optode-scalp coupling, as determined in real time by the acquisition software (Oxysoft, Artinis Medical Systems BV, the Netherlands). A three-dimensional digitizer (Patriot, Polhemus Inc., Colchester, Vermont, United States) was used to record the position. The 20 long channels [10 per hemisphere, as shown in Fig. 1(b)] were utilized to measure changes in hemoglobin, which serves as a proxy measure of brain activity, across bilateral motor regions of the brain [Fig. 1(c)], and the two short channels were used to minimize the effects of superficial hemodynamics.^26,27

Fig. 1

(a) Photographs of optode array placement on one participant’s head. (b) Registered probe geometry: yellow dots indicate the 10 sources, and cyan dots specify the eight detectors. Black numbers in green circles indicate long separation channels, and the black lines denote SSCs. (c) Sensitivity profile displayed in a log10 scale.

2.4.

Experimental Protocol 1: Fine Motor Task

A finger-tapping exercise activated the primary motor cortex (M1), and its task-related hemodynamic activity was measured. Next, subjects were told to tap their thumbs consecutively with the other fingers of the same hand (i.e., thumb and index finger tapping, thumb and middle finger tapping, etc.). The action was performed in brief bursts of 10 s, interspersed with rest periods ranging from 20 to 24 s. This variable interstimulus period was intended to prevent task events from becoming phase-locked to physiological hemodynamic oscillations²⁸ and to reduce anticipatory effects.²⁹ In addition, video cues were used to instruct movement initiation and offset. The motor task lasted around 11 min and was divided into 20 blocks (10 with the left hand and 10 with the right hand, randomly allocated), as depicted in Fig. 2(a).

Fig. 2

(a) Visualization of the experimental protocol, which includes two types of finger-tapping tasks: left and right. The entire experiment took 11 min. (b) Flow chart of the processing steps used in this study.

2.5.

Experimental Protocol 2: A 2-min Walk

The participants initially conducted a 2-min walking exercise with a baseline standing period of 20 s. The walking response was analyzed as a single trial from $-2$ to 25 s, but 2 min were allocated to see if the patients presented gait freeze. During the over-ground walking condition, participants walked back and forth along a 6 m straight path with 180-deg turns at either end. A research team member walked beside the participants to ensure their safety and stood by them when they turned.

2.6.

Experimental Protocol 3: RSFC

Participants were instructed to remain as motionless as possible in a seating position for 6 min in a dimly lighted space. Participants were asked to close their eyes, remain relaxed, and refrain from concentrating on anything specific. A research team member monitored participant conduct during the trial to ensure they did not doze off. In addition, after the trial, participants were asked if they had fallen asleep at any moment. None of the participants reported that they had fallen asleep.

2.7.

fNIRS Data Processing and Analysis

2.8.

Functional Connectivity

FC analysis was performed only on the time points identified as free of motion artifacts; discarding the motion-artifact segments has been shown to preserve interhemispheric FC.³² A procedure to remove spontaneous fNIRS fluctuations common to all channels was implemented using short separation channel (SSC) signal regression. First, the SSC signal $s$ was computed as the average time course of all short channels $s_i(t)$ ($n=2$ in our case): $s(t)=\frac{1}{n}\sum _i^n{y}_i(t)$. Then, the SSC signal was regressed from the measured fNIRS time-series $Y(t)$ using a Tikhonov regularized estimator:

$Y^{\prime }(t)$ generates channel-based connectivity matrices, computing the Pearson correlation $r$ between all channels. To obtain an index of intrahemispheric connectivity, channel-to-channel correlation coefficients in the left and right hemispheres, respectively, were averaged separately for each participant.⁴³ An index of interhemispheric connectivity was calculated by averaging all channels and their contralateral homologs.

2.9.

Statistical Analysis

A paired sample $t$-test was used to compare the mean change in hemoglobin during the pre-task baseline period and the task period to assess channel activation.⁴⁴ Comparisons between PD patients and controls were made through a non-parametric Wilcoxon–Mann–Whitney test for the task-based paradigms because part of the data was not normally distributed, according to a Shapiro–Wilk test.⁴⁵ Five different metrics extracted from the hemodynamic response were compared between controls and PD patients. Furthermore, relationships between fNIRS data from channels demonstrating statistically significant group differences and clinical variables (Table 1) were examined with Spearman’s correlation coefficients, and corrections for false positives were carried out using a false-discovery rate (FDR) adjustment.^46,47 To further investigate if the side of onset mediates the correlations between clinical variables and fNIRS features, all data from PD patients were stratified according to the side of onset (11 left and 9 right), and then, a partial correlation analysis was conducted to assess the relationship between HRF features and clinical variables while controlling for age and years of education.

Table 1

Demographic and clinical features of PD patients and controls. Unless otherwise indicated, the mean values are given, with standard deviations in parenthesis.

For the intra and interhemispheric FC, Pearson’s coefficients were transformed to a Fisher $z$-score using $Z(r)=\frac{1}{2}\ln [\frac{1+r}{1-r}]$ before computing the Wilcoxon–Mann–Whitney tests.⁴⁸

In all experimental paradigms, corrections for multiple comparisons within-subject and between-group were performed using the FDR adjustment.^46,47 Categorical variables were compared using the Chi-square test.

3.1.

Clinical Evaluation

Table 1 summarizes the clinical and demographic characteristics of the studied groups. No significant differences were found in age, proportion of male participants, years of education, or body mass index. However, the cognitive dysfunction of PD patients was significantly higher than that of control participants ($p=0.0002$), as shown by previous studies.^49–51 Because cognitive impairment is common in patients with PD and can occur at any stage of the disease,⁵² those with mild cognitive impairment were included in this study.

3.2.

Cortical Activation during Fine Motor Movement

Channels that showed significant hemodynamic activation in the motor cortex evoked by finger tapping of both the dominant (right) and non-dominant hand (left) are shown in Fig. 3.

Fig. 3

Paired sample $t$-test was used to compare the mean change in hemoglobin during the pre-task baseline period and the task period to assess channel activation. The color gradient indicates the $p$-value during the finger-tapping task. Channels in white demonstrate no statistically significant variations in hemoglobin levels between the pre-task baseline and task periods. (a) Oxyhemoglobin (HbO), (b) deoxyhemoglobin (HbR), and (c) total hemoglobin (HbT).

Five different features of the HRF were assessed for all hemoglobin types: peak value, time to peak (also known as peak latency), area under the HRF curve (AUC), mean value, and slope from stimulus onset to peak because studies have revealed that the peak latency and form, which vary across and within persons for each hemoglobin signal, task, and brain area, may be significant variables to research.^53,54 For example, a larger time-to-peak in the HRF means it takes longer for the HbO concentration to peak following neuronal activation.

Channel-by-channel comparisons revealed significant peak latency differences between the PD and control groups in channel 15 for right finger tapping and channel 10 for left finger tapping, as shown in Fig. 4(a) (see Table S1 in the Supplementary Material for correlation values corrected for false positives). The results depicted in Fig. 4(b) indicate that patients with PD had delayed hypoactivation in the motor cortex during the fine motor task with the dominant hand and delayed hyperactivation with the non-dominant hand, as measured by near-infrared spectroscopy. This delayed activation may explain the difficulty of patients with PD to perform bimanual movements, as revealed by Wu et al.⁵⁵ using fMRI. Furthermore, our findings are supported by previous studies that have found hypoactivation in the contralateral primary motor cortex elicited with a simple finger-tapping task in patients with a similar HY stage.⁵⁶

Fig. 4

Group comparisons. (a) A Wilcoxon–Mann–Whitney test with the FDR correction for false positives was used to detect group differences in mean oxyhemoglobin (HbO) changes during the finger-tapping task. The color gradient indicates a $p$-value of group differences. White channels indicate no statistically significant differences between the two groups. No significant differences were found in either HbR or HbT. (b) Average waveforms at channel 15 (right finger tapping) and channel 10 (left finger tapping). The gray rectangle denotes stimulus onset and duration.

3.3.

Correlation Analysis during Fine Motor Movement

Our findings suggest significant correlations among various measures of hemodynamic activity in the motor cortex using fNIRS and different cognitive and clinical variables. After considering only channels with a significant difference between groups and correcting correlations for false positives, only the activity on channel 10, corresponding to left finger tapping, showed statistically significant correlations with clinical variables. A breakdown of each of the correlations is shown in Fig. 5 (see Table 2 for correlation values corrected for false positives).

Fig. 5

Correlation analysis. (a) Correlation matrix between clinical variables and fNIRS-derived metrics. The asterisk sign (*) denotes a significant relationship ($p<0.05$). (b) Correlation between MoCA score and time-to-peak during left finger tapping (HbO). (c) Correlation between MoCA scores and time-to-peak during right finger tapping (HbO). (d) Correlation between HY scale and peak amplitude of left finger tapping (HbR). (e) Correlation between l-dopa equivalent daily dose and time-to-peak during left finger tapping and (f) correlation between MoCA scores and time-to-peak during left finger tapping. Shaded areas denote 95% confidence intervals.

Table 2

Correlation values between HRF features and clinical variables. Significant values after the FDR correction are displayed in bold.

The correlation between time to peak during left finger tapping of both HbO and HbT and MoCA is moderate ($r=-0.4522$, $r=-0.3382$) and significant ($p=0.0034$, $p=0.0328$, respectively); this stronger negative relationship between the time it takes for HbT and HbO to peak during a motor task with the non-dominant hand and cognitive performance possibly indicates that cognitive processes require optimal oxygen and blood supply to the brain⁵⁷ and impairments of cerebral oxygenation present in PD patients have an effect on cognitive performance.

The association between peak HRF during left finger tapping of HbR and HY scale is moderate ($r=-0.4622$) and significant ($p=0.0402$), indicating a negative relationship between the peak of HbR in the right hemisphere and the severity of PD symptoms. Previous works supported this finding; for instance, Junqué et al.⁵⁸ pointed out that asymmetries have been reported in PD, with left-hemisphere functions relatively more preserved than right-hemisphere ones; this suggests that the right hemisphere may play an important role in PD symptoms. In addition, although not explicitly addressing PD, Goldstein et al.⁵⁹ suggested that normal aging has a functionally asymmetric effect on the right hemisphere, which could potentially contribute to PD symptoms. However, further research is needed to confirm this potential negative relationship between the amplitude of deoxyhemoglobin in the right hemisphere and the severity of PD symptoms.

The correlation between time to peak during left finger tapping of HbT and levodopa equivalent daily dose is significantly moderate ($r=0.4439$, $p=0.0499$), suggesting a significant positive relationship between the time it takes for total hemoglobin to peak in the right hemisphere and medicament dosage. This may be explained by the effect of levodopa absorption and delivery to the brain.⁶⁰ No association was seen in disease duration, UPDRS score, or FOGQ.

The correlation between AUC during right finger tapping and the HY scale was found to be statistically significant: $r(17)=-0.9419$, $p=0.0454$, indicating a strong negative association between the two variables, as shown in Figs. 5(f)–5(g). The correlation coefficient of −0.9419 suggests that, after accounting for age and education, an increase in the HY scale is associated with a decrease in the AUC of the HRF in patients with right-side onset but not in patients with left-side onset. The same results were obtained with the mean value of the HRF during right finger tapping and the HY scale [Fig. 5(h)]. This supports a neural inefficiency hypothesis in which patients with more disability show a decrease in the activity evoked by finger tapping with the dominant hand. This result may be explained by the limited sample size and the fact that all patients with an HY of 4 are patients with right-side onset. Our results contrast with the findings of Zhu et al.,⁶¹ in which left-dominant symptom patients showed more severe gait impairments as well as more severe sleep-related dysfunction.

3.4.

Motor Cortex Activation during Walking

There was a significant increase in activity in the primary motor cortex for the simple walking task compared with baseline in both the control group (four channels, $t(19)=2.93$, $p<0.01$) and the PD group (one channel, $t(19)=2.46$, $p<0.03$), as depicted in Fig. 6. All active channels were found in the left hemisphere. However, there were no significant differences between patients with PD and controls; this could be explained due to two findings. First, normal walking does not significantly increase blood oxygenation in the PFC,⁶² and although it is expected in the motor cortex, it was not observed in our data. We hypothesize this may be due to technical limitations of the fNIRS probe used, such as suboptimal optode separation. Second, many studies reported that increased cortical activity during walking has confounding effects due to motion artifacts.⁶³ Further research is required to determine its causes.

Fig. 6

Paired sample $t$-test was used to compare the mean change in hemoglobin during the pre-task baseline period and the walking task period to assess channel activation. The color gradient indicates the $p$-value during the walking task. Channels in white demonstrate no statistically significant variations in hemoglobin levels between the pre-task baseline and task periods.

3.5.

FC Analysis during Walking

Several factors might explain the lack of significant differences in FC observed in the study comparing patients with PD and the control group during a walking task (Fig. 7). The inherent variability in task execution and the heterogeneity of the condition could result in increased variability in observed FC patterns, potentially masking any subtle differences.

Fig. 7

FC during walking: (a)–(c) color-coded average FC matrices (Fisher’s $z$-score) between channels for both groups calculated for HbO, HbR, and HbT, respectively. Warm colors denote positive correlations and cold colors denote negative correlations. (d)–(f) Statistical comparison between controls and PD patients. Only statistically significant correlations are shown ($p<0.05$). After the FDR correction, no significant differences were found. (g)–(i) Group comparison of interhemispheric connectivity computed for HbO, HbR, and HbT, respectively. (j)–(l) Group comparison of intrahemispheric connectivity computed for HbO, HbR, and HbT, respectively.

3.6.

RSFC Analysis

We hypothesized that there might be differences in RSFC between PD patients and control participants. The $z$-score transformed correlation matrices are shown in Figs. 8(a)–8(c). Our analysis revealed significant differences in several channels, particularly in the HbR contrast, as depicted in Figs. 8(d)–8(f). In addition, a general trend of decreased positive connections in the PD group was observed. However, these results did not survive the FDR correction. This trend of decreased connectivity in PD patients confirms observations made in previous studies.⁶⁴

Fig. 8

FC: (a)–(c) Color-coded average FC matrices (Fisher’s $z$-score) between channels for both groups calculated for HbO, HbR, and HbT, respectively. Warm colors denote positive correlations, and cold colors denote negative correlations. (d)–(f) Statistical comparison between controls and PD patients. Only statistically significant correlations are shown ($p<0.05$). After the FDR correction, no significant differences were found. (g)–(i) Group comparison of interhemispheric connectivity computed for HbO, HbR, and HbT, respectively. A statistically significant decrease in interhemispheric connectivity in PD patients compared with controls was found in HbR measurements. (j)–(l) Group comparison of intrahemispheric connectivity computed for HbO, HbR, and HbT, respectively.

Our analysis revealed significant differences between the two groups regarding the strength of interhemispheric connectivity, as shown in Figs. 8(g)–8(i). More specifically, a statistically significant decrease in interhemispheric connectivity in PD patients compared with controls was found in HbR measurements [see Fig. 8(h)]. The existing literature studying brain networks has associated a decreased interhemispheric FC with symptoms of PD. Gan et al.^65,66 found that PD patients with impulse control disorders had abnormal interhemispheric RSFC, and those patients with levodopa-induced dyskinesia had altered interhemispheric connectivity. The study found that PD patients with freezing of gait had decreased interhemispheric connectivity in certain brain regions compared with PD patients without freezing of gait and healthy controls.⁶⁷ These findings imply the possibility of using decreased interhemispheric FC as one of many potential neuroimaging biomarkers for diagnosing PD, albeit it is shared by other neurodegenerative conditions such as Alzheimer’s disease.⁶⁸

Our intrahemispheric connectivity analysis revealed trends between the groups, but these results did not pass the FDR threshold. Specifically, control participants showed stronger positive connections within a hemisphere than PD patients (see Figs. 8(j)–8(l)].