structural Brain connectome and cognitive impairment

This copy is for personal use only. To order printed copies, contact [email protected]

Sebastiano Galantucci, MD Federica Agosta, MD, PhD Elka Stefanova, MD Silvia Basaia, MSc Martijn P. van den Heuvel, PhD Tanja Stojkovic´, MD Elisa Canu, PhD Iva Stankovic´, MD Vladana Spica, MD Massimiliano Copetti, PhD Delia Gagliardi, MD Vladimir S. Kostic´, MD Massimo Filippi, MD

Purpose:

To investigate the structural brain connectome in patients with Parkinson disease (PD) and mild cognitive impairment (MCI) and in patients with PD without MCI.

Materials and Methods:

This prospective study was approved by the local ethics committees, and written informed consent was obtained from all subjects prior to enrollment. The individual structural brain connectome of 170 patients with PD (54 with MCI, 116 without MCI) and 41 healthy control subjects was obtained by using deterministic diffusion-tensor tractography. A network-based statistic was used to assess structural connectivity differences among groups.

Results:

Patients with PD and MCI had global network alterations when compared with both control subjects and patients with PD without MCI (range, P = .004 to P = .048). Relative to control subjects, patients with PD and MCI had a large basal ganglia and frontoparietal network with decreased fractional anisotropy (FA) in the right hemisphere and a subnetwork with increased mean diffusivity (MD) involving similar regions bilaterally (P , .01). When compared with patients with PD without MCI, those with PD and MCI had a network with decreased FA, including basal ganglia and frontotemporoparietal regions bilaterally (P , .05). Similar findings were obtained by adjusting for motor disability (P , .05, permutation-corrected P = .06). At P , .01, patients with PD and MCI did not show network alterations relative to patients with PD without MCI. Network FA and MD values were used to differentiate patients with PD and MCI from healthy control subjects and patients with PD without MCI with fair to good accuracy (cross-validated area under the receiver operating characteristic curve [principal + secondary connected components] range, 0.75–0.85).

Conclusion:

A disruption of structural connections between brain areas forming a network contributes to determine an altered information integration and organization and thus cognitive deficits in patients with PD. These results provide novel information concerning the structural substrates of MCI in patients with PD and may offer markers that can be used to differentiate between patients with PD and MCI and patients with PD without MCI.

1

From the Neuroimaging Research Unit (S.G., F.A., S.B., E.C., D.G., M.F.) and Department of Neurology (M.F.), Institute of Experimental Neurology, Division of Neuroscience, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Via Olgettina 60, 20132 Milan, Italy; Clinic of Neurology, Faculty of Medicine, University of Belgrade, Belgrade, Serbia (E.S., T.S., I.S., V.S., V.S.K.); Department of Psychiatry, Rudolf Magnus Institute of Neuroscience, University Medical Center Utrecht, Utrecht, the Netherlands (M.P.v.d.H.); and Biostatistics Unit, IRCCS-Ospedale Casa Sollievo della Sofferenza, San Giovanni Rotondo, Foggia, Italy (M.C.). Received February 3, 2016; revision requested April 19; revision received August 3; accepted May 25; final version accepted September 1. Address correspondence to M.F. (e-mail: [email protected] ). Supported by Ministry of Education and Science, Republic of Serbia (ON175095) and Ministero della Salute (GR091577482).

RSNA, 2016

q

Online supplemental material is available for this article.

RSNA, 2016

q

Radiology: Volume 283: Number 2—May 2017 n radiology.rsna.org

515

Original Research n Neuroradiology

Structural Brain Connectome and Cognitive Impairment in Parkinson Disease1

NEURORADIOLOGY: Structural Brain Connectome and Cognitive Impairment

P

arkinson disease (PD) is a neurodegenerative disorder characterized by a progressive decline in both motor function and cognitive function (1). Mild cognitive impairment (MCI) can be identified in approximately 25% of patients with a new diagnosis of PD (2), and patients with MCI progress to dementia more frequently than do those with normal cognitive performance (3). The neural substrates of cognitive decline in patients with PD are not yet fully elucidated (4). Dysfunction

Advances in Knowledge nn The study of the structural brain connectome showed that patients with Parkinson disease (PD) and mild cognitive impairment (MCI) had a large basal ganglia and frontoparietal network, with decreased fractional anisotropy (FA) in the right hemisphere and a subnetwork with an increased mean diffusivity (MD) bilaterally (P , .01) relative to control subjects. nn When compared with patients with PD without MCI, those with PD and MCI had a network with decreased FA, including basal ganglia and frontotemporoparietal regions bilaterally (P , .05), and similar findings were obtained by adjusting for motor disability (P , .05, permutationcorrected P = .06). nn Patients with PD without MCI did not show regional network alterations relative to healthy control subjects (P , .01, P , .05). nn Receiver operating characteristic (ROC) curve analysis with leaveone-out cross validation showed that network FA and MD values enabled differentiation of patients with PD and MCI from healthy control subjects and from patients with PD without MCI with fair to good accuracy (crossvalidated area under the ROC principal + secondary connected components range, 0.75–0.85). 516

Galantucci et al

of dopaminergic and nondopaminergic neurotransmitter systems, limbic and neocortical Lewy body disease, and Alzheimer disease–related changes are considered the main molecular and pathologic substrates of cognitive impairment in patients with PD (4). More recently, accumulating pieces of evidence from neuroimaging studies suggest an abnormal network architecture involving several cortical and subcortical systems in patients with PD and cognitive decline. To date, only a few studies have assessed the pattern of microstructural alterations of brain white matter (WM) connections in patients with PD and MCI, and these studies have suggested that cognitive deficits are associated with abnormalities of frontal and interhemispheric WM connections (5–8). More importantly, several studies showed that WM abnormalities, as revealed by diffusiontensor (DT) magnetic resonance (MR) imaging, may precede gray matter atrophy in patients with PD but without dementia (5,6). Furthermore, evidence of functional connectivity alterations in brain cognitive networks in patients with PD and MCI is emerging. By using resting-state functional MR imaging, researchers have observed a reduced connectivity between the dorsal attention network and right frontoinsular regions (9) and a functional disconnection of the frontoparietal network (10) in these patients. Conceptualization of the human brain as an integrated network (ie, the connectome) has become increasingly prevalent, and neuroimaging research is now focusing on the study of integrated models of brain function and structure rather than on individual brain areas (11). Graph theoretical analyses and network sciences form

Implication for Patient Care nn This study provides information concerning the structural substrates of MCI in patients with PD and may offer markers to differentiate patients with PD and MCI from those with PD but no impairment.

a powerful framework with which to measure the functional and structural brain organization (11). Brain networks have been described as graphs in which anatomic brain regions are defined as nodes linked by edges (ie, functional or structural connections) (11). These techniques can enable one to make a detailed description on how the disease alters brain organization (12) and, more recently, they have allowed researchers to make hypotheses about underlying physiopathology, possibly leading to identification of prognostic markers (13). Only two studies so far have investigated brain networks using graph analysis applied on resting-state functional MR imaging data in patients with PD (14,15), suggesting decreased global and nodal functional efficiency relative to those in healthy control subjects. One study indicated that the topologic properties of brain functional networks are severely impaired in patients with PD and MCI (14). To our knowledge, there have been no studies to address whole-brain structural network alterations in patients with PD and MCI by using graph analysis.

Published online before print 10.1148/radiol.2016160274 Content code: Radiology 2017; 283:515–525 Abbreviations: AUC = area under the ROC curve DT = diffusion tensor FA = fractional anisotropy MCI = mild cognitive impairment MD = mean diffusivity PD = Parkinson disease ROC = receiver operating characteristic UPDRS = Unified Parkinson Disease Rating Scale WM = white matter Author contributions: Guarantors of integrity of entire study, E.S., M.F.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, S.G., F.A., E.S., T.S., M.C., V.S.K.; clinical studies, E.S., T.S., I.S., V.S., V.S.K.; statistical analysis, S.G., S.B., M.P.v.d.H., E.C., M.C., M.F.; and manuscript editing, S.G., F.A., E.S., S.B., M.P.v.d.H., E.C., M.C., V.S.K., M.F. Conflicts of interest are listed at the end of this article.

radiology.rsna.org n Radiology: Volume 283: Number 2—May 2017


The main aim of this study was to explore WM microstructural alterations and brain network abnormalities related to cognition in a large population of patients with PD without dementia by using DT MR imaging–based structural connectomics.

Materials and Methods This prospective study was approved by the local ethics committee on human studies, and written informed consent was obtained from all subjects (or their legal guardians) prior to enrollment. Study participants were consecutively recruited from January 2011 to October 2013.

Subjects A total of 170 patients with idiopathic PD but not dementia (16) and 41 matched healthy control subjects were enrolled at the Clinic of Neurology, Faculty of Medicine, University of Belgrade. The diagnosis of PD was made by trained observers (V.S.K., T.S., I.S., V.S.; 35, 8, 5, and 8 years of experience in clinical neurology, respectively) in accordance with international criteria (16). Patients were excluded if they had parkin, leucine-rich repeat kinase 2, and glucocerebrosidase gene mutations; moderate-to-severe limb or head rest tremor; depression according to the structured clinical interview for Diagnostic and Statistical Manual of Mental Disorders, 4th edition axis I disorders; dementia according to the Movement Disorder Society diagnostic criteria for PD with dementia (17); cerebrovascular disorders or intracranial masses; a history of traumatic brain injury; or any other major neurologic or medical condition. Eleven patients were excluded because of progress to dementia, and four patients were excluded due to both dementia and cerebrovascular disease. Healthy control subjects with no history of neurologic, psychiatric, or other major medical illnesses were recruited among friends and spouses of patients and by word of mouth. All subjects underwent a comprehensive evaluation, including neurologic history taking and examination

(V.S.K., T.S., I.S., V.S) and neuropsychologic assessment (E.S., 25 years of experience in neuropsychology). Appendix E1 (online) contains full details on clinical and cognitive evaluations.

MR Imaging Study MR images were obtained with the 1.5T Philips Achieva system (Philips Medical Systems, Best, the Netherlands). Dual-echo fast spin-echo, three-dimensional T1-weighted transient field echo, and pulsed gradient spin-echo singleshot echo-planar DT MR imaging sequences were performed. See Appendix E1 (online) for complete sequence parameters. The complete MR image analysis procedure is described in Appendix E1 (online). Analyses were performed by two observers (S.G., S.B.; 8 and 2 years of experience in neuroimaging, respectively). Briefly, T1-weighted images were processed by using the Freesurfer suite (version 5.3; http://surfer. nmr.mgh.harvard.edu/), resulting in an 83-area atlas, which was used to define the brain nodes for the network analysis (Table E1 [online]). The preprocessing of DT MR imaging data was performed with FSL software (http:// www.fmrib.ox.ac.uk/fsl/). Deterministic DT tractography was perfomed with the Diffusion Toolkit and Trackvis (https://www.nitrc.org/projects/trackvis) by using the Fiber Assignment by Continuous Tracking (or FACT) algorithm (18). Tract-based spatial statistics software (version 1.2; http://fsl. fmrib.ox.ac.uk/fsl/fslwiki/TBSS) was used to perform multisubject voxelwise DT MR imaging analysis, as described previously (5). Network construction.—A brain network can be described as a graph, where the nodes are brain regions and the edges represent their connections. Here, the nodes were the 83 gray matter areas segmented with Freesurfer software, and the edges were represented by the WM tracts obtained by using the Diffusion Toolkit linking each pair of nodes (Table E1 [online]). An individual brain network was obtained for each subject included in the study (19) (Appendix E1 [online]).


Galantucci et al

Network analysis.—Global topologic network properties were obtained for each group by using the Brain Connectivity Toolbox (https://sites.google. com/site/bctnet/). The following global network metrics were explored: smallworldness (20) and mean network degree, clustering coefficient, characteristic path length, betweenness centrality, assortativity, global efficiency, and network density (number of existing connections divided by the number of possible connections). A Wilcoxon rank sum test was used to compare these metrics between groups. The results of such tests were identical for fractional anisotropy (FA) and mean diffusivity (MD) matrices for degree, characteristic path length, and density, as such network metrics only account for the presence or absence of a connection. The network-based statistic (21) was used to identify brain network differences between groups. The analysis was performed on network FA and MD data to account for the structural organization of the tracts in terms of both directionality and presence of barriers to water diffusion. Network-based statistics were used to compare FA and MD network data between groups at two levels of significance (P , .01, P , .05) (Appendix E1 [online]). The between-patient group comparison was repeated, including the Unified Parkinson Disease Rating Scale (UPDRS) III score as a nuisance variable (to adjust for disease severity). We also checked for differences in global number of streamlines, FA, and MD across groups by using the t test. Average FA and MD of the principal and smaller connected components were computed for each contrast and were entered into a receiver operating characteristic (ROC) curve analysis with leave-one-out cross validation to quantify the ability of the components to enable discrimination between groups (M.C., 8 years of experience in biostatistics).

Results Fifty-four (32%) patients with PD were classified as having PD with MCI, while the remaining 116 (68%) patients with 517

518

…

103 61 5 522 6 425.4 (0–1930)

… …

69 95

… …

163 7

12.4 6 2.6 (8–20) 57.2 6 9.1 (31–76) 5.1 6 5.2 (1–26) 28.8 6 16.1 (5–76) 43.5 6 21.5 (7–102) 1.7 6 0.8 (1–4) [94/43/32/1]

13.5 6 2.9 (8–18) … … … … …

… …

162 … 100 70 62 6 8 (39–83)

41 … 15 26 63 6 8 (49–77)

All Patients with PD (n = 170)

… … … …

… …

… …

.01 … … … … …

.37 .01 … … .68

P Value*

31 21 1 690.5 6 433.8 (0–1560)

52 2

23 29

10.9 6 2.4 (8–16) 58.2 6 9.3 (38–76) 6.2 6 4.9 (1–22) 37.2 6 16.3 (12–76) 55.8 6 21.9 (16–102) 2.1 6 0.9 (1–4) [19/17/17/1]

52 … 29 25 64 6 9 (39–81)

Patients with PD and MCI (n = 54)

35 17 2 447.4 6 356.4 (0–1200)

51 3

22 30

11.8 6 2.2 (8–17) 58.7 6 8.0 (44–74) 4.6 6 4.4 (1–19) 26.3 6 14 (7–61) 39.1 6 18.4 (11–86) 1.6 6 0.8 (1–3) [32/15/7/0]

51 … 29 25 63 6 7 (47–83)

Matched Patients with PD without MCI (n = 54)

72 40 4 443.6 6 399.6 (0–1930)

111 5

46 66

13.1 6 2.4 (8–20) 56.8 6 9.2 (31–74) 5.4 6 5.4 (1–26) 24.9 6 14.4 (5–61) 37.9 6 18.9 (7–86) 1.7 6 1 (1–3) [75/26/15/0]

110 … 71 45 61 6 8 (43–83)

All Patients with PD without MCI (n = 116)

… … … … … … …

… …

,.001 … … … … …

.46 .1 … … .48

P Value for Patients with PD and MCI vs Control Subjects

… … … … … … …

… …

.001 … … … … …

0.13 .1 … … 0.94

P Value for Matched Patients with PD without MCI vs Control Subjects

… … … … … … …

… …

.19 … … … … …

0.14 .01 … … 0.33

P Value for All Patients with PD without MCI vs Control Subjects

†

Data in brackets are numbers of patients with a score of 1–1.5, 2–2.5, 3–3.5, or 4, respectively.

* All patients with PD versus healthy control subjects.

.004

.61

.65

.98

.15 .89 .06 ,.001 ,.001 .01

.37 ..99 … … .39

P Value for Patients with PD and MCI vs Matched Patients with PD without MCI

Note.—Data are mean 6 standard deviation or number. Data in parentheses are the range. P values were obtained with analysis of variance, followed by posthoc pairwise comparisons. LEDD = Levodopa Equivalent Daily Dose,.

Right-handed Sex Male Female Age at MR imaging (y) Education (y) Age at onset (y) Disease duration (y) UPDRS III UPDRS total Hoehn and Yahr scale† Motor phenotype Tremor dominant Rigid akinetic Symmetry Asymmetric Symmetric Side of onset Right Left Symmetric LEDD

Characteristic

Healthy Control Subjects (n = 41)

Demographic and Clinical Findings in Patients with PD and Healthy Control Subjects

Table 1

NEURORADIOLOGY: Structural Brain Connectome and Cognitive Impairment Galantucci et al



Galantucci et al

Table 2 Global Topologic Network Properties of Patients with PD and Healthy Control Subjects

Connectivity Matrix Type Betweenness centrality FA MD Degree Clustering coefficient FA MD Assortativity FA MD Density Global efficiency FA MD Path length Small worldness FA MD

P Values for Matched Patients with PD without MCI vs Control Subjects

P Values for Patients with PD and MCI vs Matched Patients with PD without MCI

.03 .01 .02

.05 .11 .16

.62 .11 .16

0.25 6 0.01 0.24 6 0.01 0.25 6 0.01 7.11 31024 6 3.02 31025 7.37 31024 6 4.33 31025 7.18 31024 6 3.86 31025

.02 .002

.60 .52

.05 .02

0.13 6 0.04 0.06 6 0.04 0.1298 6 0.004

.02 .06 .02

.61 .03 .16

.005 .03 .16

.35 .67 .14

.01 .08 .16

.67 .22

.48 .35

Control Subjects

Patients with PD with MCI

Patients with PD without MCI

145.88 6 6.34 138.82 6 4.82 9.65 6 0.37

149.68 6 7.88 142.02 6 5.61 9.44 6 0.38

148.64 6 7.37 140.48 6 5.68 9.54 6 0.37

0.15 6 0.05 0.07 6 0.04 0.1273 6 0.004

0.13 6 0.05 0.05 6 0.04 0.1286 6 0.005

P Value for Patients with PD and MCI vs Control Subjects

0.1642 6 0.006 0.1578 6 0.010 0.1616 6 0.008 ,.001 4.17 31024 6 1.91 31025 4.25 31024 6 2.44 31025 4.16 31024 6 1.68 31025 .12 2.66 6 0.06 2.69 6 0.07 2.68 6 0.07 .01 5.05 6 0.78 5.13 6 0.54

5.17 6 0.65 5.33 6 0.60

5.06 6 0.57 5.29 6 0.66

.21 .08

Note.—Data are mean 6 standard deviation. P values refer to Wilcoxon rank-sum test models.

PD were found to be without cognitive impairment. A subgroup of 54 patients with PD without MCI was matched with a group of patients with PD with MCI according to demographic features and was selected to perform the betweenpatient group comparison, holding the same statistical power of the PD MCI group (Table 1, Table E2 [online]).

Structural Brain Networks: Global Network Properties The results of the analysis of global topologic network properties are reported in Table 2. When compared with control subjects, patients with PD without MCI showed increased betweenness centrality in the FA matrices (P = .05) and decreased assortativity in the MD matrices (P = .03). When compared with control subjects, patients with PD and MCI showed increased betweenness centrality (P = .03 for FA matrices, P = .01 for MD matrices), characteristic

path length (P = .01), and assortativity (P = .02 for FA matrices). Degree (P = .02), clustering coefficient (P = .02 for FA matrices, P = .002 for MD matrices), density (P = .02), and global efficiency (P , .001 for FA matrices) were decreased. When compared with patients with PD without MCI, patients with PD and MCI showed increased assortativity (P = .005 for FA matrices, P = .03 for MD matrices) and reduced clustering coefficient (P = .05 for FA matrices, P = .02 for MD matrices) and global efficiency (P = .01 for FA matrices). There were no significant differences in small worldness among the different groups.

Structural Brain Networks: Networkbased Statistic Between-group comparisons.—The networks of affected structural connectivity in patients with PD and MCI relative to control subjects and patients with PD without MCI are shown in Figure 1.


This figure displays both the principal connected component (ie, the largest subnetwork of affected connectivity) and the smaller subnetworks with altered FA or MD values. All the P values obtained from the network analysis and reported here are permutation-corrected P values. When compared with control subjects, patients with PD without MCI showed a decreased global number of streamlines (P = .03), while no global FA (P = .27) or MD (P = .43) alterations or regional structural brain network abnormalities were found. In addition, there was no network impairment when the 54 patients with PD without MCI who were matched with patients with PD and MCI were compared with healthy subjects (at P , .05, corrected P = .35 for FA connections and corrected P = .45 for MD connections; at P , .01, corrected P = .77 for FA networks and corrected P = .58 for MD networks). 519


Galantucci et al

Figure 1

Figure 1: Affected structural connections in patients with PD and MCI relative to healthy control subjects and patients with PD without MCI (networkbased statistic). The principal connected component is shown in red, and the other affected connections not included in the principal connected component are shown in green. Table E1 (online) reports the names of each brain node with the corresponding number. (a) Subnetworks show altered structural connectivity (decreased FA or increased MD) in patients with PD and MCI relative to healthy control subjects at P , .01 (Fig 1 continues).

When compared with control subjects, patients with PD and MCI showed a decreased global number of streamlines (P = .048) and FA (P = .005) and an increased global MD (P = .004). The network analysis of patients with PD and MCI versus control subjects at P , .01 (Fig 1a) showed a basal ganglia and frontoparietal network (or principal connected component) with decreased FA that included the following regions in the right hemisphere: putamen, thalamus, frontal pole, rostral middle frontal gyrus, inferior frontal gyrus-pars orbitalis, lateral orbitofrontal cortex, insula, and inferior parietal gyrus (P , .001). Smaller secondary networks with reduced FA were found in the left hemisphere connecting the thalamus, temporal pole, and fusiform gyrus (P = .04) (Fig 1a). Right middle temporal and superior parietal gyri (P = .02), as well as the right paracentral gyrus and brainstem, were also part of these smaller 520

networks (P = .02). Relative to control subjects, patients with PD and MCI had a large subnetwork with increased MD that included the following regions: the pallidum bilaterally; the right caudate, putamen, thalamus, orbital and lateral regions of the frontal lobe; and the superior temporal lobe (P , .001) (Fig 1a). Patients with PD and MCI also had numerous smaller subnetworks with increased MD compared with control subjects, involving the left basal ganglia and precentral gyrus, insula, and several regions of the temporal, parietal, and occipital lobes bilaterally (P , .001 to P , .01) (Fig 1a). When P , .05 was adopted (see Network Analysis section and Appendix E1 [online] for further details), patients with PD and MCI had more extensive altered principal networks with decreased FA or increased MD relative to control subjects (Fig 1b). The impaired FA subnetwork included connections

linking the bilateral caudate and thalamus, left globus pallidus, right frontotemporoparietal regions, and left temporoparietooccipital regions (P = .01) (Fig 1b). The network with increased MD expanded even more and included the majority of the connections of both hemispheres (P = .002) (Fig 1b). At the global level, patients with PD and MCI showed reduced FA (P = .03) and increased MD (P = .02) without number of streamlines (P = .34) changes compared with matched patients with PD without MCI (P = .03). The network analysis at P , .01 showed no difference between patients with PD and MCI and matched patients with PD without MCI (P = .20 for FA connections, P = .17 for MD connections). When patients with PD and MCI were compared with matched patients with PD without MCI at P , .05 (Fig 1c), a principal network of affected FA connections was found, including the left



Galantucci et al

Figure 1 (continued)

Figure 1 (continued). (b) Subnetworks show altered structural connectivity (decreased FA or increased MD) in patients with PD and MCI relative to healthy control subjects at P = .05. (c) Subnetworks show impaired structural connectivity (decreased FA) in patients with PD with MCI relative to matched patients with PD without MCI at P = .05.

thalamus and right basal ganglia and the frontal, temporal, and parietal regions (P = .04), with a pattern similar to that seen in patients with PD and MCI compared with control subjects.

We did not find affected MD connections in patients with PD and MCI compared with matched patients with PD without MCI (P = .06). When UPDRS III was introduced as a nuisance


variable, the results were similar, albeit at a less significant threshold (P , .05, permutation-corrected P = .06; Fig E1 [online]). ROC curve analysis.—Table 3 shows results of the ROC curve analysis with leave-one-out cross validation. Average FA and MD values of the (combined) principal and secondary components enabled us to differentiate patients with PD and MCI from healthy control subjects with fair to good accuracy (crossvalidated area under the ROC curve [AUC] range, 0.75–0.85). Network FA demonstrated good AUC values in distinguishing patients with PD and MCI from those with PD without MCI (principal component + secondary components = 0.81).

Voxelwise Analysis In patients with PD with MCI compared with control subjects, a decrease in FA 521


Galantucci et al

Table 3 ROC Curve Analysis with a Leave-One-Out Cross Validation to Differentiate PD Patients with Mild Cognitive Impairment from Healthy Control Subjects and Patients without Cognitive Deficits Group and DT MR Imaging Metric

P Value Threshold

Component

Cross-validated AUC

Cross-validated AUC*

Patients with PD and MCI vs Control Subjects FA FA FA FA MD MD MD MD

.01 .01 .05 .05 .01 .01 .05 .05

FA FA

.05 .05

Principal 0.80 0.82 Secondary 0.83 … Principal 0.85 0.85 Secondary 0.71 … Principal 0.74 0.78 Secondary 0.79 … Principal 0.76 0.75 Secondary 0.65 … Patients with PD and MCI vs Patients without MCI Principal 0.74 0.81 Secondary 0.77 …

* Data are principal + secondary connected components.

was found in bilateral frontal, parietal, and temporal WM, involving the genu and anterior part of the corpus callosum and associative frontoparietal tracts, with a slight prevalence on the right side (Fig 2). The pattern of MD increase was similar with additional involvement of the midbrain and pontine WM. An MD increase that was less extensive but had a similar spatial pattern was found when matched patients with PD without MCI were compared with control subjects. The FA reduction in this last comparison was confined to part of the right superior frontoparietal associative tracts. There was no significant difference when patients with PD and MCI were compared with matched patients with PD without MCI.

Discussion This study evaluated the structural DT MR imaging connectome correlates of MCI in a large sample of patients with PD but without dementia. The main finding of our study was that PD with MCI is associated with a structural network disruption involving basal ganglia and the frontotemporoparietal region. This finding agrees with the clinical and cognitive aspects of the 522

disease characterized by motor, executive, and visuospatial deficits. Despite the fact that both patients with PD and MCI and patients with PD without MCI showed microstructural damage measured by using tract-based spatial statistics, only patients with PD and MCI had significant alterations at the network level when compared with control subjects and patients with PD without MCI. When looking at global topologic network features, patients with PD and MCI had severe alterations when compared with control subjects, while patients with PD without MCI showed very few abnormalities. These results suggest that cognitive impairment in patients with PD is likely the consequence of a disruption of complex structural brain networks rather than degeneration of individual WM bundles. Such network alterations enabled us to distinguish patients with PD and MCI from control subjects and patients with PD without MCI with a high level of accuracy. A few previous studies have explored the pattern of DT MR imaging alterations in patients with PD and MCI with voxelwise or tractography-based approaches (5–7). The pattern of DT MR imaging changes in patients with

PD and MCI was similar to the pattern we described in our sample using tractbased spatial statistics. In addition, we found a somewhat similar pattern of MD increase in patients with PD without MCI relative to control subjects, while the significant FA decrease was much less distributed and mostly confined to part of the right superior frontoparietal associative tracts. Previous results agree with this finding and showed that alterations of DT MR imaging metrics can be found throughout the brain at the earliest stages of PD, even before the occurrence of cognitive deficits (22,23). Reduced FA values in patients with PD compared with control subjects have been described in the frontal lobes, without corresponding gray matter volume loss (24), involving the WM adjacent to the supplemental motor areas, prefrontal areas, and anterior cingulate gyrus (24). DT MR imaging changes in patients with PD also have been found in the genu of the corpus callosum and superior longitudinal fasciculus (25). More recently, DT MR imaging abnormalities were detected in the nigrostriatal fibers (26). The analysis of structural connectome that we performed deepened the findings of voxelwise analyses (including the one performed in our sample). Whereas voxelwise approaches detect diffusion changes at the voxel level, the structural connectome analysis considers the relationships between degenerating connections and is able to provide pieces of information on sets of connections that are likely degenerating simultaneously (12). On the basis of our results and those in the literature, we hypothesize that WM microstructural alterations are not sufficient to cause cognitive impairment; rather, a severe disruption of structural connections between brain areas forming a network is required to determine altered information integration and organization and thus cognitive deficits. We are aware that the overall severity of the disease may have contributed to the network disruption independently of the cognitive status. Even if we are not able to completely exclude the influence of disease severity, it is worth noting that the pattern of



Galantucci et al

Figure 2

Figure 2: MR images show microstructural WM damage (tract-based spatial statistics). Areas of decreased FA (red) and increased MD (blue) overlapped on the overall FA skeleton (green) in groups of patients with PD compared with healthy control subjects (P , .05 corrected for multiple comparisons).

differences was confirmed when adjusting for UPDRS III score. Moreover, we found no network abnormalities, even when we compared all patients with PD without MCI to control subjects, thus decreasing the effect size required to meet statistical significance. The association between brain network modifications and cognitive impairment in patients with PD and MCI has been described previously in a functional connectome study (14). Altered global network parameters of clustering, modularity, and small-worldness correlated with deficits in visuospatial

and visuoperceptual abilities, as well as memory functions in patients with PD and MCI (14). The involvement of the frontoparietal and frontoinsular networks and their functional disconnection as the substrate of MCI in patients with PD have also been suggested by other resting-state functional MR imaging studies using independent component analysis (9,10). The network alterations we found in our study are possibly the structural correlates of functional connectivity derangement observed in previous populations with PD and MCI (9,10,14).


This study was not without limitations. The first limitation, which is common to all DT MR imaging studies reflecting the limitations of the DT model, was the presence of crossing fibers, which can cause uncertainty in the estimation of the tensor model. Thus, potential errors in tract reconstruction with DT tractography may have occurred in regions where such fiber disposition is more frequent. Second, the reference standard for the regional parcellation of brain MR imaging has not yet been established. Thus, the choice of network nodes 523


was somewhat arbitrary across studies and was based on either anatomic templates or data-driven parcellations (27). We were aware that the Freesurfer method does not take into account the cytoarchitectonic differences that can be present in different parts of the same parcellated area. However, even considering such a limitation, this segmentation enables a reliable definition of anatomic regions using a robust and automated method and has already been used in previous connectome studies (19). Third, motor deficits were more severe; consequently, levodopa equivalent doses were higher in patients with PD and MCI than in patients with PD without MCI, even when a demographically matched sample was selected. It should be noted that the presence of MCI in patients with PD is inherently related to more severe motor deficits (1). Thus, selecting a population of patients with PD without MCI matched for motor performance with patients with PD with MCI was not possible, despite the fact that we recruited a large cohort of patients with PD. However, it is worth noting that when we repeated the analysis adjusting for UPDRS III, the pattern of differences between patients with PD and MCI and patients with PD without MCI did not change. We chose not to control the analysis additionally for levodopa equivalent doses because we expected it to be highly correlated to UPDRS III, as demonstrated in our sample (UPDRS III vs levodopa equivalent doses: Spearman correlation coefficient = 0.71, P , .001). In conclusion, our results suggest that whole-brain connectome analysis with DT MR imaging may be a useful tool with which to explore the structural abnormalities related to a cognitive decline in PD. This study indicates that MCI in patients with PD is associated with decreased FA and increased MD in structural networks connecting frontotemporoparietal areas and basal ganglia. On the contrary, PD without MCI is associated with frontoparietal WM microstructural alterations but not structural network disruption. The distribution and 524

Galantucci et al

pattern of the structural connectome alterations we found suggest that such an approach might enable identification of novel markers of PD-related cognitive impairment. This could be particularly valuable if, in the future, novel disease-modifying drugs specific to those patients at high risk of developing cognitive impairment are developed. Moreover, the improvement of our ability to predict cognitive impairment would also be helpful to improve a patient’s awareness about cognitive impairment and dementia. Future longitudinal studies are warranted to confirm these findings and to describe the progression of structural network abnormalities over time in this condition and how they correlate with or can be used to predict functional network abnormalities and clinical outcome. Disclosures of Conflicts of Interest: S.G. disclosed no relevant relationships. F.A. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: received speaker honoraria from Excemed. Other relationships: disclosed no relevant relationships. E.S. disclosed no relevant relationships. S.B. disclosed no relevant relationships. M.P.v.d.H. disclosed no relevant relationships. T.S. disclosed no relevant relationships. E.C. disclosed no relevant relationships. I.S. disclosed no relevant relationships. V.S. disclosed no relevant relationships. M.C. disclosed no relevant relationships. D.G. disclosed no relevant relationships. V.S.K. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: received grants from Stada, Valeant, Boehringer Ingelheim, and Novartis. Other relationships: disclosed no relevant relationships. M.F. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: provided consulting services, served as a speaker, or both for Teva Pharmaceutical Industries, Biogen Idec, Excemed, and Novartis; institution received money from Biogen Idec, Novartis, and Teva Pharmaceutical Industries. Other relationships: disclosed no relevant relationships.

References 1. Barone P, Aarsland D, Burn D, Emre M, Kulisevsky J, Weintraub D. Cognitive impairment in nondemented Parkinson’s disease. Mov Disord 2011;26(14):2483–2495. 2. Muslimovic D, Post B, Speelman JD, Schmand B. Cognitive profile of patients with newly diagnosed Parkinson disease. Neurology 2005;65(8):1239–1245.

3. Broeders M, de Bie RMA, Velseboer DC, Speelman JD, Muslimovic D, Schmand B. Evolution of mild cognitive impairment in Parkinson disease. Neurology 2013;81(4):346–352. 4. Halliday GM, Leverenz JB, Schneider JS, Adler CH. The neurobiological basis of cognitive impairment in Parkinson’s disease. Mov Disord 2014;29(5):634–650. 5. Agosta F, Canu E, Stefanova E, et al. Mild cognitive impairment in Parkinson’s disease is associated with a distributed pattern of brain white matter damage. Hum Brain Mapp 2014;35(5):1921–1929. 6. Hattori T, Orimo S, Aoki S, et al. Cognitive status correlates with white matter alteration in Parkinson’s disease. Hum Brain Mapp 2012;33(3):727–739. 7. Matsui H, Nishinaka K, Oda M, et al. Wisconsin Card Sorting Test in Parkinson’s disease: diffusion tensor imaging. Acta Neurol Scand 2007;116(2):108–112. 8. Rae CL, Correia MM, Altena E, Hughes LE, Barker RA, Rowe JB. White matter pathology in Parkinson’s disease: the effect of imaging protocol differences and relevance to executive function. Neuroimage 2012;62(3):1675–1684. 9. Baggio HC, Segura B, Sala-Llonch R, et al. Cognitive impairment and resting-state network connectivity in Parkinson’s disease. Hum Brain Mapp 2015;36(1):199–212. 10. Amboni M, Tessitore A, Esposito F, et al. Resting-state functional connectivity associated with mild cognitive impairment in Parkinson’s disease. J Neurol 2015;262(2):425– 434. 11. Rubinov M, Sporns O. Complex network measures of brain connectivity: uses and interpretations. Neuroimage 2010;52(3): 1059–1069. 12. Crossley NA, Mechelli A, Scott J, et al. The hubs of the human connectome are generally implicated in the anatomy of brain disorders. Brain 2014;137(Pt 8):2382–2395. 13. Fornito A, Zalesky A, Breakspear M. The connectomics of brain disorders. Nat Rev Neurosci 2015;16(3):159–172. 14. Baggio HC, Sala-Llonch R, Segura B, et al. Functional brain networks and cognitive deficits in Parkinson’s disease. Hum Brain Mapp 2014;35(9):4620–4634. 15. Skidmore F, Korenkevych D, Liu Y, He G, Bullmore E, Pardalos PM. Connectivity brain networks based on wavelet correlation analysis in Parkinson fMRI data. Neurosci Lett 2011;499(1):47–51. 16. Daniel SE, Lees AJ. Parkinson’s Disease Society Brain Bank, London: overview and



research. J Neural Transm Suppl 1993;39: 165–172. 17. Emre M, Aarsland D, Brown R, et al. Clinical diagnostic criteria for dementia associated with Parkinson’s disease. Mov Disord 2007; 22(12):1689–1707; quiz 1837 . 18. Mori S, Kaufmann WE, Davatzikos C, et al. Imaging cortical association tracts in the human brain using diffusion-tensor-based axonal tracking. Magn Reson Med 2002;47(2): 215–223. 19. Verstraete E, Veldink JH, Mandl RC, van den Berg LH, van den Heuvel MP. Impaired structural motor connectome in amyotrophic lateral sclerosis. PLoS One 2011;6(9): e24239. 20. Watts DJ, Strogatz SH. Collective dynamics of ‘small-world’ networks. Nature 1998; 393(6684):440–442.

21. Zalesky A, Fornito A, Bullmore ET. Network-based statistic: identifying differences in brain networks. Neuroimage 2010;53(4): 1197–1207. 22. Ziegler E, Rouillard M, André E, et al. Mapping track density changes in nigrostriatal and extranigral pathways in Parkinson’s disease. Neuroimage 2014;99:498–508. 23. Meijer FJ, Bloem BR, Mahlknecht P, Seppi K, Goraj B. Update on diffusion MRI in Parkinson’s disease and atypical parkinsonism. J Neurol Sci 2013;332(1-2):21–29.

Galantucci et al

study. AJNR Am J Neuroradiol 2009;30(6): 1222–1226. 26. Zhang Y, Wu IW, Buckley S, et al. Diffusion tensor imaging of the nigrostriatal fibers in Parkinson’s disease. Mov Disord 2015;30(9): 1229–1236. 27. Zalesky A, Fornito A, Harding IH, et al. Wholebrain anatomical networks: does the choice of nodes matter? Neuroimage 2010;50(3): 970–983.

24. Karagulle Kendi AT, Lehericy S, Luciana M, Ugurbil K, Tuite P. Altered diffusion in the frontal lobe in Parkinson disease. AJNR Am J Neuroradiol 2008;29(3):501–505. 25. Gattellaro G, Minati L, Grisoli M, et al. White matter involvement in idiopathic Parkinson disease: a diffusion tensor imaging


525