Early functional network alterations predict motor and cognitive decline in parkinson’s disease

Abstract

This study aimed to identify potential structural and resting state FC (rs-FC) alterations in de novo PD and examine their possible relationship with motor and cognitive symptoms; and to explore whether early structural and rs-FC alterations were associated with subsequent clinical progression. Seventy-eight de novo PD patients and thirty-one healthy subjects (HS) were enrolled. The severity of motor symptoms was assessed by the MDS-Unified Parkinson’s Disease rating scale and cognitive performance was assessed with the Montreal Cognitive Assessment. Forty out of 78 PD patients underwent a clinical follow-up after a period of 5.29 ± 1.40 years. All participants underwent 3T MRI scanning. Structural MRI analyses included gray matter volume estimation, thalamus, basal ganglia and cerebellar volumetry. Resting-state functional connectivity was assessed using independent component analysis. Compared to HS, de novo PD patients did not show any structural alteration. Conversely, they showed decreased rs-FC in several brain networks, including the default mode network, sensorimotor, cerebellar, medial visual, occipital, orbitofrontal, dorsal attention, executive control, and the left frontoparietal networks. Although baseline functional alterations were not associated with clinical measures at the time of assessment, reduced baseline rs-FC in most resting-state networks was predictive of clinical progression over time. De novo PD patients exhibit widespread rs-FC alterations across multiple brain networks, despite a preserved structural integrity. Early rs-FC alterations in sensorimotor, cerebellar, and cognitive networks were linked to subsequent clinical deterioration. These findings highlight the potential of rs-fMRI as a valuable early imaging correlate for tracking disease progression in PD.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-34136-7.

Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by a complex interplay of pathophysiological processes, primarily affecting the nigrostriatal pathway and subsequently involving other brain structures^1,2. Structural imaging techniques have been employed to investigate disease progression in PD³, with several studies reporting changes such as gray matter atrophy, iron accumulation, and depigmentation of the substantia nigra as potential biomarkers^4–6. However, these findings remain to be consistently validated^7,8.

Resting-state functional MRI (rs-fMRI) has also been used to explore the pathophysiological mechanisms underlying PD, as well as the neural correlates of both motor and non-motor symptoms^9–12. Functional alterations have been observed in both the basal ganglia-thalamo-cortical and cerebello-thalamo-cortical circuits^13–16. In particular, in the early stage of the disease altered functional connectivity in some cerebellar regions (such as fastigium, dentate and interposed nuclei) have been described and appeared to be related to motor impairment¹³. In addition, other studies in de novo and moderate PD patients provided evidence of functional alterations between subcortical nuclei such as the subtalamic nucleus and regions within the sensorimotor cortex¹⁵. Finally, PD patients in advanced disease stage and disabling motor symptoms such as freezing of gait frequently exhibit abnormal FC mainly in the corticopontine-cerebellar pathways (in the bilateral cerebellum and in the pons), as well as in the the executive (frontal gyrus and in the angular gyrus) and visual areas (occipito-temporal gyrus)^14,16.

Additional evidence has highlighted disruptions in higher-order networks such as the frontoparietal (dorso-laterale and prefrontal cortex)¹⁷, dorsal attention (intraparietal sulcus)¹⁸, and default mode networks (right medial temporal lobe and bilateral inferior parietal cortex)¹⁹ in particular in PD patients with impairment of cognitive fuctions.

Nonetheless, most resting-state functional connectivity (rs-FC) alterations have been reported in patients with advanced PD ²⁰, while only a few studies have described functional changes in early-stage, drug-naïve patients^{13,16,21–23}, suggesting that rs-FC alterations may serve as early indicators of neurodegeneration^20,24.

Ideally, neuroimaging biomarkers should be capable of predicting disease progression at early stages, prior to significant neurodegeneration, and should be validated against objective measures of clinical decline over a sufficiently long follow-up period. To date, no study has comprehensively met these criteria. Specifically, no study has examined early-stage rs-FC and structural alterations and subsequently assessed their predictive value for motor and cognitive progression.

In the present study, we aimed to fill this research gap by investigating a large cohort of de novo PD patients with disease duration of less than two years and longitudinal clinical follow-up of at least 3.5 years. We first identified potential structural and rs-FC alterations in early PD and assessed their association with motor and cognitive symptoms. The main aim of the study, in a subgroup of patients with follow-up of approximately five years, was to explore whether early imaging alterations could predict clinical progression. This approach may offer novel insights into PD pathophysiology and contribute to the identification of reliable early-stage imaging correlate for disease monitoring.

Materials and methods

Study design and patients

Seventy-eight de novo PD patients and thirty-one age- and sex-matched healthy subjects (HS) were enrolled.

Inclusion criteria were: (i) age over 18 years, (ii) diagnosis of PD according to international criteria²⁵; (iii) clinical history less than two years (iii), and (iv) neurological examination by expert neurologists. Exclusion criteria were: (i) other neurological or psychiatric diagnoses, (ii) moderate to advanced PD (Hoehn and Yahr stage III-V), (iii) prior levodopa or other dopaminergic treatments at the baseline, (iv) presence of dyskinesia (v), cognitive impairment as assessed by a Montreal Cognitive Assessment (MoCA) score of ≤ 26. None of the HS received treatment with drugs known to induce parkinsonism or were related to a case patient involved in the study. A movement disorder specialist (SP or DB) performed the clinical assessment. The severity of motor symptoms was assessed by the MDS Unified Parkinson’s Disease rating scale (MDS-UPDRS) III²⁶, whereas cognitive performance was assessed with the MoCA²⁷ by the same physicians.

Forty out of 78 PD patients underwent a clinical follow-up after a period of 5.3 ± 1.5 years. Differences in UPDRS III and MoCA scores between the two visits were calculated as delta (Δ) = (follow-up score − baseline score) / baseline score. A score of 26 or less at MOCA was considered as an indicator of cognitive impairment²⁸. Demographic and clinical parameters of this PD subgroup are reported in Table 1. Clinical scales were uniformly administered by the same physicians at both timepoints.

Table 1.

Demographic and clinical characteristics of de Novo PD patients at follow-up.

	Baseline N = 40	Follow-up N = 40	Δ
Age (years)	61.68 ± 10.65	65.63 ± 10.42	3.85
Gender (F/M)	10/30	10/30	-
Disease duration (years)	1.19 ± 0.60	5.29 ± 1.40	4.05
UPDRS III	15.66 ± 6.12	26.39 ± 10.66	0.83
MOCA	28.81 ± 1.37	25.80 ± 3.25	-0.10
LEDD	-	388 ± 133
H&Y [median (range)]	1 (1–5)	1.5 (1–5)	-

F: females; M: males; UPDRS-III: Unified Parkinson’s Disease Rating Scale, part III; MoCA: Montreal Cognitive Assessment; LEDD: levodopa equivalent daily dosage; H&Y: Hoehn and Yahr Scale. Values are expressed as mean ± SD.

Policlinico Umberto I Research Ethics Committee (Sapienza University of Rome) approved the study. All participants gave their written informed consent before participating in the study, which was conducted following the latest revision of the Declaration of Helsinki.

MRI data acquisition

Images were acquired with a 3-Tesla scanner (Siemens Magnetom Verio, Erlangen, Germany) and a 12-channel head coil designed for parallel imaging (GRAPPA). All the patients were drug naïve (hence they were scanned in the OFF state). The following sequences were acquired:

• High-resolution three-dimensional (3D) T1-weighted (T1-3D) MPRAGE sequence (repetition time (TR) = 2,400 ms, echo time (TE) = 2.12ms, inversion time (TI) = 1000 ms, flip angle = 8◦, field of view (FOV) = 256 mm, matrix = 256 × 256, 176 sagittal slices 1-mm thick, no gap);

• Blood oxygen level-dependent (BOLD) single-shot echo-planar imaging (TR = 3,000 ms, TE = 30 ms, flip angle = 89°, FOV = 192 mm, 64 × 64 matrix, 140 volumes, voxel size = 3mm3), with all patients instructed to close their eyes and remain awake during the rs-fMRI acquisitions.

• Dual turbo spin-echo, proton density (PD) and T2-weighted images (TR = 3320 ms, TE1 = 10ms, TE2 = 103 ms, FOV = 220 mm, matrix = 384 × 384, 25 axial slices 4-mm thick, 30% gap).

• High-resolution 3D fluid-attenuated inversion recovery (FLAIR) sequence (TR = 6000 ms, TE = 395 ms, TI = 2100 ms, FOV = 256 mm, matrix = 256 × 256, 176 sagittal slices 1-mm thick, no gap).

An expert radiologist (PP) examined all MRIs, in particular T2-weighted and FLAIR images, to exclude the presence of concomitant brain abnormalities and focal white matter hyperintensities.

MRI data analysis

Anatomical and functional preprocessing were performed using fMRIPrep 20.2.3²⁹ (RRID: SCR_016216), which is based on Nipype 1.7.0³⁰ (RRID: SCR_002502) and FSL v6.0.4 (https://fsl.fmrib.ox.ac.uk/fsl/docs/#/).

Structural MRI measures

3D-T1 MR images of PD patients and HS were segmented into gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) using the Computational Anatomy Toolbox (CAT12, http://dbm.neuro.uni-jena.de/cat/), an extension toolbox of Statistical Parametric Mapping software (SPM12, http://www.fil.ion.ucl.ac.uk/spm/software/spm12). For each subject, total GM, WM and CSF volumes were obtained and normalized for head size using the total intracranial volume (TIV).

Deep GM structures, i.e., the thalamus, caudate nucleus, putamen, and globus pallidus, were segmented from T1-3D images using FMRIB’s Integrated Registration and Segmentation Tool (FIRST, http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST). For each deep GM structure, the left and right volumes were summed, and normalized for head size using TIV.

Cerebellar structures were identified, and volumes were calculated using the SUIT toolbox³¹. Cerebellar structures included 13 bilateral lobules and 8 vermis regions. Each subject’s cerebellum was isolated and cropped from the T1-3D anatomical images. Each cropped image was subsequently normalized into SUIT space using generated flow field and affine transformations. Lastly, the probabilistic cerebellar atlas was resliced back into the individual subject space to identify cerebellar structures. We carefully inspected SUIT outputs for each subject to ensure accuracy. The calculated cerebellar volumes were normalized to the TIV. For further analysis, global cerebellar volume was used.

Functional connectivity

Independent component analysis (ICA) of preprocessed functional data was performed using FSL’s MELODIC tool (Multivariate Exploratory Linear Optimized Decomposition into Independent Components)³². For group-wise ICA, a single 4D data set was created by temporally concatenating preprocessed functional data. The dimensionality of group-ICA was assessed using different numbers of components (i.e., 20, 25, 30, 35, 40). Lastly, a dimensionality of 35 was chosen, as the explained data variance was sufficient to obtain good estimates of the signals and well-known rs-networks (RSNs) were identified³³. RSNs of interest covered the entire brain and were selected via spatial correlation coefficients (fslcc tool) using RSNs generated by Smith et al.³³ and Yeo et al.³⁴ templates, and then verified by expert visual inspection (C.P., 10 years of experience and S.P. 12 years of experience).

The set of spatial maps from the group-average analysis was used to generate subject-specific versions of the spatial maps and associated time series using a dual regression technique^35,36. For each subject, the group-average set of spatial maps was first regressed (as spatial regressors in a multiple regression) into the subject’s 4D space-time dataset, resulting in a set of subject-specific time series, one per group-level spatial map. These time series were then regressed (as temporal regressors in a multiple regression) into the same 4D dataset, resulting in a set of subject-specific spatial maps, one per group-level spatial map.

Statistical analyses

Statistical analyses of demographic, clinical and structural MRI parameters were performed using SPSS statistics software (version 22.0). The Shapiro–Wilk normality test was performed to check for the normal distribution of demographic and clinical data. Parametric and non-parametric tests were used for normally and non-normally distributed data, respectively. Between-group differences were tested using Mann–Whitney U-test and Chi-square test for continuous and dichotomous variables, respectively (p < 0.05 for null hypothesis rejection).

Spearman correlation analyses were performed between structural MRI measures and baseline clinical scores (UPDRS III and MoCA).

In the PD patients who underwent the clinical follow-up, we performed multiple regression analyses using structural MRI measures (cerebral cortex, thalamus, caudate, putamen, pallidum and cerebellum volumes) as independent variables and ΔUPDRS III or ΔMOCA as dependent variables. Age and sex were inserted as covariates of no-interest.

We have also compared baseline clinical, structural and functional MRI data between the 40 PD de novo patients who underwent the clinical follow up and the 38 remaining PD de novo patients to exclude significant differences in clinical and radiological features between the two groups (see Supplementary Table 1).

Within-network rs-FC

Functional subject-specific spatial maps obtained from dual regression were entered into group-level voxel-wise analyses. For each RSN, we investigate rs-FC differences between PD patients and HS using a two-sample unpaired t-test. Age, sex, and TIV were used as covariate of no interest in all analyses. Voxel-wise statistical analyses were performed with permutation-based non-parametric statistics using the FSL Randomise permutation-based program with 5,000 permutations³⁷.

The Randomise tool (5000 permutations) was also used to perform voxel-wise general linear model (GLM) analyses between within-network rs-FC and clinical features (baseline UPDRS III and MoCA scores).

For those PD patients having clinical follow-up, we also performed voxel-wise GLM analyses between within-network rs-FC and clinical changes (ΔUPDRS III and ΔMoCA scores).

All results were corrected using false discovery rate (FDR) correction³⁸ for multiple comparisons (p < 0.05). Anatomical localization of significant clusters was established according to the Harvard-Oxford Cortical Structural Atlas included in FSL (http://www.fmrib.ox.ac.uk/fsl/data/atlas descriptions.html).

Between-network rs-FC

Between-network rs-FC differences between PD patients and HS were investigated using the FSLNets toolbox (http://fsl.fmrib.ox.ac.uk/fsl/fslwki/FSLNets). After normalization of the extracted time courses of all RSNs identified in each subject, time courses of artificial components and components of no-interest were regressed out of the individual data. Subject-wise correlation matrices of both full and partial correlation of all remaining RSNs time courses were then created. The resulting correlation coefficients were then Fisher z-transformed and corrected for temporal autocorrelation. Between-subject testing was then performed across correlation values acquired for pairs of independent components. Between-group comparisons of time series correlations were performed using non-parametric unpaired t-test (FSL Randomise tool, 5000 permutations; age, sex, TIV and LEDD as covariates of no-interest). Results were corrected using FDR correction for multiple comparisons (p < 0.05).

Results

Descriptive statistics for demographic and clinical parameters in the entire group of de novo PD and HS are reported in Table 2. There were no significant differences in age (p = 0.09) or sex (p = 0.65) between de novo PD patients and HS. In the patients, the mean UPDRS score was 16.76 ± 8.54 and the mean MoCA score was 27.47 ± 2.21.

Table 2.

Demographic and clinical characteristics of healthy subjects (HS) and de Novo PD patients.

Demographic /clinical feature	HS N = 31	PD N = 78	p-value
Age (years)	64.23 ± 9.32	61.26 ± 9.35	0.09
Gender (F/M)	12/19	25/53	0.65
Disease duration (years)	-	1.64 ± 1.05	-
Age at onset (years)	-	58.72 ± 8.97	-
UPDRS III	-	16.76 ± 8.54	-
MOCA	-	27.47 ± 2.21	-
H&Y [median (range)]	-	1.5 (1–5)	-

HS: healthy subjects.; PD: patients with de novo PD; F: female; M: male; UPDRS-III: Unified Parkinson’s Disease Rating Scale, part III; MoCA: Montreal Cognitive Assessment; H&Y: Hoehn and Yahr Scale. Values are expressed as mean ± SD if not stated otherwise. Differences between groups were assessed using the Mann-Whitney U test and Chi-square test for continuous and dichotomous variables, respectively (p < 0.05).

We did not find any significant differences in clinical and structural and functional MRI data between the 40 PD de novo patients who underwent the clinical follow up and the 38 remaining PD de novo patients (Supplementary Table 1).

Demographic and clinical characteristics of de novo PD patients who underwent clinical follow-up are reported in Table 1.

Structural MRI measures

GM volume did not significantly differ between de novo PD patients and HS (p > 0.05) (Table 3). Similarly, the FIRST analysis showed similar basal ganglia and thalamus volumes in de novo PD patients and HS (Table 3). Likewise, SUIT analysis yielded similar cerebellar volumes in PD patients compared to HS (p > 0.05) (Table 3).

Table 3.

GM volumes in de novo PD patients and healthy subjects (HS).

	HS	PD	p-value
Gray matter	39.05 ± 3.61	38.90 ± 3.01	0.83
Cerebellum	7.63 ± 0.67	7.81 ± 0.54	0.96
Thalamus	0.93 ± 0.06	0.95 ± 0.07	0.98
Caudate nucleus	0.43 ± 0.04	0.44 ± 0.06	0.88
Pallidum	0.22 ± 0.03	0.24 ± 0.03	0.92
Putamen	0.58 ± 0.05	0.58 ± 0.07	0.85

HS: healthy subjects.; PD: patients with de novo PD. Raw gray matter, cerebellum, thalamus, caudate, pallidum and putamen nuclei volumes were normalized within each subject as a ratio of intracranial volume and reported as a fraction (%). Differences between groups were assessed using the Mann-Whitney U test, FDR corrected for multiple comparisons.

Resting state functional connectivity

ICA yielded 35 independent components representing group-averaged networks of brain regions which temporally correlated BOLD fMRI signals. Of these, we identified 12 components with the highest spatial correlation coefficients with RSNs templates: default mode, left and right frontoparietal, dorsal attention, executive control, lateral and medial visual, orbitofrontal, occipital, cerebellar, basal ganglia and sensorimotor networks.

Within-network rs-FC

In comparison to HS, de novo PD patients showed lower rs-FC in the sensorimotor, cerebellar, medial visual, occipital, orbitofrontal, dorsal attention, executive control, and left frontoparietal networks (Fig. 1, Supplementary Table 2). Conversely, PD patients showed higher rs-FC in the default mode network, and clusters of both higher and lower rs-FC in the right frontoparietal network (Fig. 1, Supplementary Table 2). No significant differences were found between PD and HS in the basal ganglia network.

Fig. 1.

RSNs showing significant functional connectivity differences between healthy subjects (HS) and PD patients (p < 0.05, FDR corrected). Results for each RSN are overlaid onto the corresponding network (green) in the MNI152 standard brain. Red-yellow and blue-light blue color bars represent t values.

Between-network rs-FC

No significant differences in partial or full correlations between RSNs were found between de novo PD patients and HS after correction for multiple comparisons.

MRI-clinical associations

We did not find any significant association between structural MRI measures and either baseline or delta clinical scores.

We did not find any significant association between within-network rs-FC and clinical scores at baseline. On the other hand, we found negative associations between within-network rs-FC of regions belonging to the sensorimotor and cerebellar networks and the Δ UPDRS III scores, indicating that the lower the rs-FC the greater the motor worsening (Fig. 2, Supplementary Table 3). Similarly, we found positive associations between within-network rs-FC of several regions of the default mode, left frontoparietal, executive, dorsal attention, and cerebellar networks and the Δ MoCA scores, indicating that the lower the rs-FC the greater the cognitive worsening (Fig. 3, Supplementary Table 3).

Fig. 2.

Significant negative associations between within-network rs-FC of SMN and CRB and Δ UPDRS III scores (p < 0.05, FDR corrected). Results for each RSN are overlaid onto the corresponding network (green) in the MNI152 standard brain. The blue-light blue color bars represent t values.

Fig. 3.

Significant positive associations between within-network rs-FC of DMN, FPL, ECN, DAN, CBN and Δ MoCA scores (p < 0.05, FDR corrected). Results for each RSN are overlaid onto the corresponding network (green) in the MNI152 standard brain. The red-yellow color bars represent t values.

Discussion

In the present study conducted in de novo PD patients, we observed that rs-FC was significantly altered within multiple brain networks, predominantly showing decreased connectivity. Notably, these functional alterations occurred in the absence of volumetric loss in key gray matter structures. Although baseline rs-FC changes were not directly correlated with clinical measures at the time of assessment, lower baseline rs-FC across most networks was significantly associated with greater motor and cognitive progression over time.

Structural and functional connectivity in early PD

As also described by other authors, despite the absence of significant volumetric loss in the cerebral cortex, basal ganglia, thalamus, and cerebellum^39,40, de novo PD patients exhibited decreased rs-FC within several intrinsic brain networks. These included the left and right frontoparietal, dorsal attention, executive control, medial visual, occipital, cerebellar, and sensorimotor networks. In contrast, increased within-network rs-FC was observed only in the default mode network and, to a lesser extent, in the right frontoparietal network.

Only a limited number of previous studies have investigated rs-FC in newly diagnosed, drug-naïve PD patients. Consistent with our findings, these studies also reported reduced within-network rs-FC in the frontoparietal, visual, and sensorimotor networks^21,22,41, as well as in the cerebellum¹³. In PD, decreased connectivity within these networks has been linked to cognitive dysfunction, particularly in the more advanced stages of the disease^10,17,42,43, although similar alterations have also been reported in early-stage PD ⁴⁴. Notably, reduced rs-FC in the frontoparietal and dorsal attention networks has also been documented in cognitively preserved PD patients^42,44,45, suggesting that these changes may precede clinical cognitive decline.

In addition to cognitive networks, we also observed reduced rs-FC in the visual and occipital networks⁴⁶, as well as in sensorimotor and cerebellar networks^13,47,48.

Unlike most previous studies^19,49–51, we found increased rs-FC within the default mode network. Although decreased connectivity is typically associated with neuronal dysfunction, increased FC can also have pathological or compensatory significance. One interpretation is that this pattern reflects an early adaptive response in cognitively intact patients^20,52,53. This is also consistent with findings of reduced default mode network connectivity in more advanced PD, which is often associated with cognitive decline^43,50,51.

Consistent with this view, in our cohort higher baseline default rs-FC was associated with better cognitive outcomes at longitudinal follow-up (see below paragraph). This may align with reports of reduced default mode network connectivity in more advanced PD, often linked to cognitive decline^43,50,51. Nevertheless, this interpretation remains speculative in the absence of stratification by cognitive phenotypes.

Interestingly, we did not observe significant rs-FC alterations within the basal ganglia network. This contrasts with findings in more advanced PD and suggests that, in the early stages of the disease, basal ganglia function may be preserved or compensated for by other brain regions. Such compensation may delay the onset of functional disruptions in this network^22,54–56.

Furthermore, unlike the clear alterations observed in within-network connectivity, we found no significant differences in between-network rs-FC between PD patients and healthy controls. Previous studies have shown reduced inter-network coupling, particularly among networks implicated in cognitive processing, in patients with cognitive impairment²⁰. We speculate that disrupted inter-network communication emerges later in the disease course, possibly in association with cognitive decline, and may not yet be detectable in the earliest stages of PD when cognition remains preserved.

Longitudinal associations with motor and cognitive decline

The key novelty of this study lies in evaluating whether early alterations in rs-FC can predict subsequent worsening of motor and cognitive functions in de novo PD patients. Longitudinal clinical assessments revealed that early reductions in rs-FC within the sensorimotor and cerebellar networks were significantly associated with greater motor deterioration, as measured by changes in the UPDRS-III score. These findings are consistent with the established role of the sensorimotor and cerebellar networks in motor control^56,57, suggesting that their early dysfunction may contribute to the progressive motor impairments characteristic of PD.

Similarly, rs-FC reductions in regions belonging to the left frontoparietal, executive control, and cerebellar networks were associated with greater cognitive decline, as indicated by longitudinal changes in MoCA scores. Notably, higher baseline rs-FC in the default mode network was positively associated with better cognitive outcomes over the 5-year follow-up. Increased activity in this network involved the posterior component and in particular the precuneus and the posterior cingulate cortex, suggesting that FC alterations in these areas reflect adaptive process aimed at preserving cognitive function during early disease stages. This hyperconnectivity may help offset declining connectivity in other networks, such as the executive control network, which showed a negative association with cognitive performance over time.

Specifically, reduced rs-FC within the executive control network was significantly associated with MoCA decline, indicating that early connectivity disruptions in this network may contribute to worsening executive dysfunction as a hallmark of PD-related cognitive impairment.

Moreover, positive associations between baseline rs-FC in the dorsal attention and left frontoparietal networks and MoCA outcomes further support the role of these networks in cognitive resilience.

Notably, the observed association between reduced cerebellar rs-FC and worsening MoCA scores highlights the cerebellum’s emerging role in the pathophysiology of cognitive disturbances in PD. This finding reinforces evidence from previous studies^13,58 suggesting that cerebellar disconnection may contribute to cognitive deficits even in the earliest stages of the disease. Finally, baseline rs-FC measures were not associated with concurrent clinical performance, despite their predictive value for future decline. This apparent discrepancy may indicate that network-level alterations precede clinical manifestations, reflecting an early phase of functional reorganization that is not yet detectable with standard clinical scales in a quite preserved clinical population. In this sense, early FC alterations may represent a “preclinical” indicator of vulnerability, becoming clinically relevant only as the disease progresses and network inefficiency increases. These findings support the interpretation of rs-FC as an early imaging correlate of future decline.

Limitations

Several limitations of this study must be acknowledged. First, the relatively small sample size of the clinical follow-up group may have limited the statistical power to detect more subtle associations. However, this patient subgroup did not differ in clinical and MRI characteristics from patients who did not undergo follow-up and may be considered representative of the entire de novo PD patient group. Further studies on larger and independent patient samples are required to confirm our results. Second, the HS group did not receive a full neuropsychological assessment or longitudinal follow-up. This limitation may have limited our ability to fully exclude the influence of subclinical factors in the control group. Third, while we focused on de novo PD patients without dopaminergic treatment at baseline, the long-term clinical progression of PD may have been affected by pharmacological and/or physical therapy that were not fully considered in data analysis. Furthermore, our statistical analyses have been conducted with the awareness of the potential impact of the unbalanced sample sizes on our results. To address this concern, we employed appropriate statistical tests designed for unequal sample sizes, such as non-parametric unpaired t-tests. Additionally, we have applied statistical models that allow for the adjustment of potential confounding variables (age, sex, GM volume), further enhancing the precision of our analyses. Our sample balance is in line with several previous studies that privileged a higher number of patients than healthy controls, often with a ratio of 2:1 in PD de novo patients^59–62. Furthermore, we also acknowledge that mixed-effects models or slope estimation would provide a more robust characterization of longitudinal progression, particularly in the presence of non-linear disease trajectories. However, these approaches typically require at least three measurement timepoints per subject, whereas our dataset included only two. Moreover, although resting-state fMRI using ICA is a valuable tool for evaluating brain network functional activity, it is limited by its inability to capture patterns of functional dynamics of overall functional architecture of the brain that may provide further insight into disease mechanisms. Finally, we correlate only baseline fMRI data with clinical outcomes after several years, but without repeated scans it is not possible to determine whether altered connectivity is a stable imaging correlate or whether it evolves dynamically over time.

Conclusions and clinical implications

In conclusion we report that early-stage, drug-naive PD patients exhibit widespread alterations in rs-FC, particularly within motor- and cognition-related networks, in the absence of structural changes.

Notably, decreased connectivity in sensorimotor, cerebellar, and cognitive networks can predict a greater motor and cognitive decline over time suggesting that rs-fMRI may be an early imaging correlate of PD progression.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by SP, CP and DB. The first draft of the manuscript was written by SP, CP and DB and all other authors commented on previous versions of the manuscript and were contributors in writing the manuscript. All authors read and approved the final manuscript.

Funding

Authors received no funding for this study. Progetto Ricerca Corrente funded at INM Neuromed by the Italian Ministry of Health.

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.