Alterations in topological and dynamical parameters correlate with disease biomarkers and neuropsychological scores in prodromic stages of dementia

Alterations in topological and dynamical parameters correlate with disease biomarkers and neuropsychological scores in prodromic stages of dementia | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Alterations in topological and dynamical parameters correlate with disease biomarkers and neuropsychological scores in prodromic stages of dementia Anita Monteverdi, Matteo Cotta Ramusino, Francesca Conca, Sofia Manzon, and 13 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6550081/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract BACKGROUND Mild cognitive impairment (MCI) is a clinical condition in the continuum between normal cognition and dementia. Despite numerous studies, the heterogeneity of the underlying pathophysiology prevents a precise prediction of clinical evolution. METHODS In a cohort composed of MCI, healthy controls (HC), and Alzheimer’s disease (AD) patients, graph theory (GT) was combined with virtual brain modelling (TVB) to extract the information on network topology and dynamics embedded in magnetic resonance imaging (MRI) data. With this approach, the analysis was extended to a multiparametric space and brought from the group to the individual subject level. RESULTS The comparison of network properties in HC, MCI, and AD revealed a profound reshaping of brain connectivity, which mainly affected the default mode, limbic, attention, and somatosensory networks. Interestingly, positivity to AD biomarkers (Aβ and τ) in MCI correlated with network topology, while a TVB parameter (i.e., recurrent excitation) correlated with reduced global cognition (MMSE score). There was a high correlation (R 2 ~ 70%) between GT and TVB parameters and neuropsychological performance in multiple cognitive domains. CONCLUSIONS The combination of GT and TVB parameters was superior to the individual techniques alone in providing a subject-specific phenotype of MCI sensitive to molecular biomarkers and correlated with neuropsychological scores. This, in turn, could form the basis for a more precise MCI stratification leading, in the future, to a personalized prediction of evolution and therapeutic intervention. brain dynamics excitatory/Inhibitory balance mild cognitive impairment Alzheimer’s disease resting-state networks virtual brain modelling graph theory Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Background The non-invasive investigation of changes in brain networks can contribute to our understanding of clinical conditions associated with high neurological, neuropsychological and neuropathological heterogeneity, such as mild cognitive impairment (MCI). The diagnosis of MCI on the basis of core clinical criteria 1 covers a range of widely different conditions and progression modalities. A recent meta-analysis based on 89 studies with a mean 5.2 year follow-up revealed a risk of progression to dementia of 41.5% in clinical and 27.0% in population-based studies, with dementia due to Alzheimer’s disease (AD) as the most common outcome 2 . The MCI individuals remaining stable were 49.3% in clinical and 49.8% in population studies. Reversion to normal cognition was 8.7% in clinical and 28.2% in population studies. While biomarker positivity, including plasma-based measurements 3 , increased the risk of progression to dementia, a considerable variation in speed of progression and in clinical phenotype was found also in MCI due to AD 4 . While current evidence mostly addresses biomarkers of pathology and their correlation with brain structure (e.g. volume loss) and function (regional metabolism reduction due to synaptic loss), little is known about the large-scale organization of network dynamics. The structural and functional reorganization of brain network architecture with dementia progression has been assessed with graph theory (GT) measures 5 applied to resting state connectivity data, which increasingly demonstrate their sensitivity to topological alterations linked to cognitive decline 6 . However, this type of analysis cannot address the reorganization of circuit dynamics. A critical issue that waits for a solution is that the group analysis strategy, frequently used to investigate brain pathologies, gets short in capturing the multifactorial nature and the heterogeneity of subject profiles typical of the dementia syndrome. Thus, a multiparametric analysis projected in a subject-specific space can be expected to provide better insight into the link between network structural and functional alterations and large-scale brain dynamics. Here we have applied a recent technology to generate a virtual brain to simulate the neurophysiological underpinnings of network dynamics in a non-invasive and personalized way. The Virtual Brain (TVB) 7 is a data-driven modelling framework that yields information about the ensemble properties of cells and microcircuits starting from magnetic resonance imaging (MRI) data. TVB combines mesoscopic neural mass models of neural dynamics with structural MRI data to create a brain “avatar”, which is made of nodes (brain areas) and edges (axonal fiber tracts derived with tractography), to simulate large-scale brain dynamics. Thus, TVB links brain structure, function, and dynamics into a generative mechanistic model. Appropriate computational representations of local microcircuit functions enable TVB to quantify physiological parameters like NMDA and GABA receptor-mediated transmission and recurrent excitation 8 , which are especially important to address alterations of synaptic excitation and inhibition in neurodegeneration 9 , 10 . The potential of TVB to capture pathological heterogeneity and foster personalized patient profiling has been demonstrated in epilepsy 11 , stroke 12 , brain tumors 13 , multiple sclerosis 14 and advanced dementia stages 15 , 16 , and is further extended here to the case of MCI. The capacity of TVB to simulate brain dynamics can be complemented by GT that can extract, from the same MRI data, topological information about network structural and functional organization. Here, we combined TVB with GT (Fig. 1 ) to extract information from MRI data and project the parameters into a high-dimensional space. Networks supporting specific integrative and sensorimotor functions 17 , 18 [including the default mode network (DMN), limbic network (LN), attention network (AN), and somatomotor network (SMN)] were considered in a cohort including healthy controls (HC), MCI (both AD biomarker positive and negative, i.e., MCI + and MCI − ) and AD. The aim was to investigate the potential of multiparametric analysis of MRI data to yield single subject profiles correlating network mechanisms with Aβ and τ, and neuropsychological scores. Methods Subjects A total of 60 subjects were recruited at the IRCCS Mondino Foundation (Pavia, Italy): 18 HC (69 ± 5 years, 9 females), 22 MCI subjects (75 ± 6 years, 14 females), and 20 AD patients (71 ± 7 years, 16 females) (Suppl. Table 1). The informed consent was collected for all subjects and the study was approved by the local ethical committee and carried out in accordance with the Declaration of Helsinki. The diagnosis was based on current clinical criteria 1 , 19 . Etiological diagnosis of MCI was provided by cerebrospinal fluid (CSF) analysis (i.e. Aβ and τ protein levels) or amyloid PET: MCI + (10 subjects − 7A + T + by CSF analysis, 3A + by PET) and MCI − (12 subjects). The exclusion criteria were diagnosis of any neurological or psychiatric condition different from AD and secondary causes of cognitive decline such as metabolic, iatrogenic, toxic, and endocrine. Neuropsychological assessment All subjects underwent a neuropsychological examination assessing the global cognitive function with the Mini-Mental State Examination (MMSE) and a standardized battery of tests to assess different cognitive functions. Neuropsychological tests assessed the following cognitive domains: verbal short-term memory (digit span forward), verbal working memory (digit span backward), verbal long-term memory (short story test and FCSRT), visuo-spatial long-term memory (Rey–Osterrieth complex figure recall), phonological and semantic fluency (Category and Phonemic Fluency tasks), visuo-constructional function (Rey-Osterrieth complex figure copy) and visual attention and task switching (trail making test part A and B). According to the reference norms for the Italian population, raw scores for each test were corrected for age, education and sex and converted into equivalent scores (ES) ranging from 0 to 4 (0 indicating a pathological performance) 20 (Suppl. Table 1). Structural and functional connectivity MRI acquisitions were performed using a 3T Siemens Skyra scanner with a 32-channel head coil. The protocol included diffusion weighted imaging (DWI) and resting-state fMRI (rs-fMRI) scans 21 . Preprocessing of DWI and rs-fMRI data were performed according to established techniques (see Supplementary Methods). To consider both cerebro-cortical, subcortical, and cerebellar regions, an ad-hoc gray matter parcellation atlas was created combining 93 AAL and 33 SUIT labels. The gray matter nodes were then remapped onto the Buckner (cerebellar) and Yeo (cerebral) 17 , 18 atlases identifying six networks known to support integrative functions (DMN, FPN, LN, AN) and sensorimotor functions (VN, SMN) relevant to MCI and AD. DWI and rs-fMRI data was then used to construct structural and functional connectivity (SC and FC) matrices of these networks (see Supplementary Methods). Topological characterization of resting-state networks For each subject, the main GT metrics (Brain Connectivity Toolbox, Matlab) were obtained from static FC and SC matrices of resting state brain networks 5 . Parameters describing the arrangement of network nodes were the density and the core nodes. Structural and functional networks integration and segregation were evaluated with global efficiency and clustering coefficient values. Combining the path length and clustering coefficient measures allowed to define the small-worldness, which reflects the tendency of a network to balance local segregation with long-distance integration. As a measure of node centrality the betweenness centrality was computed from path lengths. For the SC matrix we further considered node strength, and shortest path length. For the FC matrix, we considered intraFC (see Supplementary Methods). Virtual Brain modelling For each subject and each network, TVB simulations were performed using the SC matrix (Fig. 1 ) and the nodes were endowed with Wong-Wang (WW) neural mass models 8 (Suppl. Figure 1). The WW is made of excitatory and inhibitory neural populations connected through NMDA synapses (J NMDA ), and GABA-A synapses (J i ), and recurrent excitation (w+). The SC matrix weights inter-node connections and is scaled by the global coupling (G), which denotes long-range coupling strength between nodes. All parameters were initially set as in Deco and colleagues 8 (Suppl. Table 2) and then tuned by model optimization. The simulated neural activity was convolved with the Ballon-Windkessel hemodynamic model to reconstruct the resting-state BOLD signal. Simulation generated an 8-minute rs-fMRI time-series per node to calculate the simulated FC and FCD matrices, which were updated iteratively (Suppl. Figure 2) against the experimental FC and FCD matrices. The metrics used for optimization were Pearson Correlation Coefficient (PCC) for FC and Kolmogorov-Smirnov distance (D KS ) for FCD. An optimal simulation required the highest PCC and the lowest D KS minimizing the cost function \(\:(1-PCC)+{D}_{KS}\) . Statistics and machine learning GT measures and TVB parameters for each network were tested for normality (Shapiro-Wilk, p < 0.05). Then, a multivariate general linear model (GLM) followed by post-hoc Bonferroni correction was applied to detect network topology and excitation/inhibition differences between groups correcting for age and gender differences. To reduce the parameter space a decision-tree-based Random Forest algorithm was applied to both GT and TVB extracting the parameters that best separated the clinical groups (AD, MCI, and HC). The top five uncorrelated features were used as input for the clustering analysis (Silhouette and the K-means tests, see Supplementary Methods). Differences between clusters were assessed using multivariate GLM. A posteriori analysis of subjects’ distributions between clusters was used to determine the correspondence between mechanistic parameters (topology and dynamics), molecular biomarker positivity (Aβ and τ), and global cognitive score (MMSE). Multiple regression analysis was performed to explore the relationship between the neuropsychological scores of specific cognitive domains and GT/TVB parameters. The backward regression algorithm automatically removed predictors until selecting the significant ones (F test, p < 0.05) explaining the neuropsychological scores variance. Results The subjects included in this study belonged to three groups: HC, MCI, and AD. All subjects underwent structural and functional MRI scans, which were used to reconstruct the structural and functional connectome and to perform GT measures and TVB simulations (Fig. 1 ). Structural, functional and dynamical characterization of resting-state networks at different dementia stages (group comparison) GT measures revealed the topological organization of resting-state networks (Fig. 2 ). From a structural point of view (Fig. 2 A), MCI patients showed higher clustering coefficient, global efficiency, and nodal strength in the DMN compared to AD (p < 0.05, Bonferroni corrected). From a functional point of view (Fig. 2 B), MCI patients showed higher values of clustering coefficient, global efficiency, and intra functional connectivity (intraFC) in the AN, VN and SMN compared to AD (p < 0.05, Bonferroni corrected). AD patients also showed lower global efficiency, clustering coefficient and intraFC in AN, VN and SMN, accompanied by higher betweenness centrality in AN compared to HC. TVB simulations yielded subject-specific parameters describing the dynamics of excitation and inhibition in resting-state networks. Model parameters related to excitation levels (J NMDA and w + ) of the DMN showed significant differences between groups (Fig. 2 C), with AD patients showing higher J NMDA and lower w + compared to HC and MCI (p < 0.05, Bonferroni corrected). In aggregate, at group level, none of the GT or TVB parameters showed significant differences between MCI and HC, while several differences emerged between HC or MCI and AD. Subject-specific profiles of resting-state network topology and dynamics Parameters describing network topology and dynamics in GT and TVB analysis were given as input to machine learning algorithms (random forest for features selection followed by k-mean clustering and Silhouette analysis) to generate multiparametric subject-specific profiles. An a posteriori analysis of subject-specific distributions between clusters was performed to evaluate the correspondence between GT and TVB parameters, Aβ and τ biomarkers (MCI + , MCI − ), and cognitive decline (MMSE score) in MCI patients. Topological characterization Topological parameters of LN, AN and SMN (Fig. 3 ) were identified by random forest as the most informative ones distinguishing pathological from healthy subjects (Suppl. Figure 3). The five most meaningful uncorrelated parameters were structural betweenness centrality of LN, node strength of LN, functional global efficiency of AN, structural global efficiency of SMN and functional small-worldness of LN. Based on these five parameters, Silhouette and K-means analysis identified three main topology-based clusters (Suppl. Figure 4). Ct1 was characterized by lower values of AN global efficiency (p < 0.001) and higher values of LN functional small-worldness compared to other clusters (p < 0.05). This cluster mainly contained AD patients. Ct2 showed higher values of LN node strength (p < 0.001) and SMN structural global efficiency compared to the other clusters (p < 0.005), while AN global efficiency was increased compared to Ct1 but decreased compared to Ct3 (p < 0.001). This cluster mainly contained MCI patients which, in turn, were mostly belonging to MCI + (8 out of 9). Ct3 showed higher values of AN global efficiency (p < 0.001) and lower values of structural LN betweenness centrality compared to the other clusters (p < 0.05). This cluster mainly contained HC subjects and MCI patients entirely belonging to MCI − (8 out of 8). Dynamic properties characterization TVB parameters of LN, AN, DMN, FPN, and SMN (Fig. 4 ) were identified by random forest as the most informative ones distinguishing pathological from healthy subjects (Suppl. Figure 5). The five most meaningful uncorrelated parameters were recurrent excitation (w + ) of LN, AN, DMN, FPN, and global coupling (G) of SMN. Based on these five parameters, Silhouette and K-means analysis identified three clusters (Suppl. Figure 6). Cd1 was characterized by higher values of w + in LN (p < 0.001) and G in SMN compared to other clusters (p < 0.05) and mainly contained HC and MCI patients. Cd2 showed lower values of w + in DMN compared to other clusters (p < 0.001) and contained most of AD patients (17 out of 21). Cd3 showed higher values of w + in the AN network compared to other clusters (p < 0.001) and mainly contained MCI patients. It should be noted that MMSE of MCI patients in Cd2 was significantly lower than MMSE of MCI patients in Cd1 or Cd3. Correspondence between multiparametric profiling, molecular biomarkers and cognitive decline It is worth recapitulating here the main correlations that emerged between parameters reported in Figs. 3 and 4 . A posteriori analysis of topology-based cluster composition showed that the majority of MCI − was in Ct3 together with the majority of HC, while MCI + mostly fell in Ct2, without a significant difference in MMSE score. The a posteriori analysis of dynamics-based cluster composition showed heterogeneity of MCI − and MCI + between clusters. However, MCI patients belonging to Cd2 showed significantly lower MMSE compared to MCI patients in other clusters. The relationship of network topology and dynamics with neuropsychology To assess the correlation between GT and TVB parameters with subject-specific clinical severity, patient-specific neuropsychological scores were used in a backward regression model against network-specific parameters. Interestingly, combinations of GT and TVB parameters were more effective than individual measures in explaining the variation of neuropsychological scores across multiple cognitive domains (Fig. 5 A), with an explained variance (adjusted R 2 ) reaching in some cases ~ 70%. Examples of these correlations for visual attention and task switching in the AN, verbal working memory in the DMN, and semantic fluency in the LN, are shown in Fig. 5 B. Discussion This work provides the first evidence that a subject-specific profiling of MCI patients can be performed through the combination of GT, which describes the topological organization of networks, and TVB, which simulates network dynamics. This approach identifies a rich set of network alterations that allows to identify clusters of MCI patients positive to AD molecular biomarkers (Aβ and τ) or sensitive to global cognitive changes (MMSE), and to correlate microscopic network parameters with neuropsychological scores. Conversely, no differences between MCI and HC did emerge at the group level (not unexpectedly, differences emerged between HC or MCI and AD, though). Thus, the GT/TVB multiparametric analysis extending over multiple network properties and cognitive domains in single patients is superior to classical low-dimensional analysis at group level for addressing the well-known heterogeneity of the clinically defined MCI condition. The topological analysis (GT) revealed a reorganization of LN, AN, and SMN. A crucial finding is that subjects were grouped into three clusters, one hosting most MCI + and another most MCI − patients, suggesting that, even if clinical severity is not different, molecular biomarker positivity is already assigning some of the MCI subjects to the prodromal phase of AD. The AN functional global efficiency decreased with the AD/HC ratio in the clusters suggesting that a decreased inter-regional communication plays a role in the cognitive decline from HC to MCI to AD. The LN functional small-worldness increased with the AD/HC ratio in the clusters suggesting a desynchronization of distant brain regions associated with an increase of short-distance synchronization. The LN higher betweenness centrality in the clusters with high AD/HC ratio may facilitate network integration possibly implying mechanisms compensating for the detrimental effect of small-worldness. The LN high node strength in the cluster with the majority of AD and MCI + patients may also play a compensatory role against synaptic pruning 22 . The SMN showed the highest structural global efficiency in the cluster containing most of MCI + suggesting that it might enhance cognitive control 23 compensating the derangement in AN and LN. In aggregate, specific changes in AN, LN, and SMN unveil a profound alteration of the integration/segregation balance in MCI along with possible compensation antagonizing the evolution to dementia. The dynamical analysis (TVB) revealed marked changes in the recurrent excitation of LN, AN, and DMN, and in the global coupling of SMN, allowing to group subjects into three clusters (distinct from those obtained with topological analysis). Here, while the segregation of MCI + and MCI − is less clear-cut, there is a significant reduction of the MMSE score in MCI subjects belonging to the cluster with the highest AD/HC ratio and the lowest values of network recurrent excitation. Consistently, an extended recurrent excitation in cortical circuits boosts local processing supporting decision making and working memory tasks 24 , which are reduced in MCI and even more in AD. The higher G value in the SMN implies an increased network synchronization 25 , a sign of synaptic dysfunction already reported in AD 15 , 16 that is now extended to MCI patients. It should also be noted that FPN, which is involved in frontotemporal dementia but not in AD, did not show significant changes of recurrent excitation, in agreement with previous studies 15 , 16 . The combination of topological and dynamical parameters (GT and TVB) explained neuropsychological scores variance (up to 70%) in multiple cognitive domains better than the individual measures alone, implying that the measures are (at least partly) independent and provide an incremental amount of information. Consistently, visual attention and task switching, semantic fluency, and verbal short-term memory correlated with parameter changes in DMN, AN, and LN, confirming a main involvement of these networks in the worsening of cognitive functions 26 . In aggregate, while classical group analysis captures changes in the AD brain compared to MCI and HC, single subject multiparametric analysis reclassifies subjects, unveiling changes in MCI that correlate with pathology biomarkers (GT) and the global cognitive state (TVB). Not unexpectedly, given the different sensitivity of the two techniques, pathology markers reflect tissue damage while brain dynamics reflect cognitive processing. The low integration/segregation balance and recurrent excitation in MCI + and AD patients places DMN, AN, and LN networks, which are responsible for memory and executive functions, at the core of the pathogenetic process. Tau PET imaging studies show that the extension of tau pathology from the temporo-mesial region towards neocortical temporal and parietal regions is a strong predictor of progression to dementia 27 . It is therefore not surprising that DMN, LN and AN, which include temporo-mesial and temporo-parietal areas, could be among the first and most affected networks in the clusters of patients with the majority of MCI + (7/10) and AD (17/20). The SMN, at odd with cognitive decline in MCI, shows enhanced global efficiency and may therefore be involved in compensatory processes, as also previously suggested for the AN 28 . This study should be seen as the beginning of a broader set of investigations and can be improved in several respects. First, although the sample size does not impact per se on the TVB ability of uncovering excitation/inhibition profiles, the study of a larger cohort is warranted. Moreover, GT and TVB analysis may be refined by curating tractography and using specific circuit models 29 , which may impact on structural and functional connectivity and brain dynamics. Specific aspect that may be considered further concern the role of the cerebellum, whose functional connectivity is markedly increased in MCI 30 , and of AN, which may better divided into the dorsal and ventral attention networks (DAN and VAN). In conclusion, the combined GT/TVB analysis allows to detect a set of changes in the DMN, LN, AN, and SMN of MCI patients presumably reflecting a combination of pathogenetic alterations and compensatory mechanisms. Then, multiparametric profiling identifies clusters of subjects with high expression of molecular biomarkers and reduced MMSE scores. In a clinical perspective, a longitudinal analysis of GT/TVB results is warranted to predict who, among the MCI patients, will evolve into AD and benefit from a timely intervention, opening new perspectives for personalized treatments in prodromal dementia stages. Declarations Data sharing statement All codes used for brain dynamics simulations with TheVirtualBrain are available as a Python code that can be found at https://www.thevirtualbrain.org/tvb/zwei. The dataset will be made available on Zenodo. Declaration of interest Authors report no conflict of interest. Acknowledgements This work was performed at the IRCCS Mondino Foundation and was supported by the Italian Ministry of Health (SG-2021-12374430) to AM. ED’A acknowledges #NEXTGENERATIONEU (NGEU) and the Ministry of University and Research (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006) – A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022) and EBRAINS-Italy (Project IR0000011, CUP B51E22000150006). CW-K acknowledges BRC (#BRC704/CAP/CGW), MRC (#MR/S026088/1), Ataxia UK, Rosetrees Trust (#PGL22/100041 and #PGL21/10079). CGWK is a shareholder in Queen Square Analytics Ltd. EL is a PhD student enrolled in the National PhD in Artificial Intelligence, XXXVIII cycle, course on Health and life sciences, organized by Università Campus Bio-Medico di Roma. MG acknowledges “National Centre for HPC, Big Data and Quantum Computing” (Project CN00000013 PNRR MUR - M4C2 - Fund 1.4 - Call “National Centers” - law decree n. 3138 16 December 2021). Funding This work was supported by the Italian Ministry of Health (SG-2021-12374430). References Albert S. Marylin, D. T. S. et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: Recommendations from the National Institute on Aging- Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011 May ; 7(3) 270–279. doi10.1016/j.jalz.2011.03.008. 7 , (2011). Salemme S, Lombardo FL, Lacorte E, et al. The prognosis of mild cognitive impairment : A systematic review and meta-analysis. Alzheimer’s Dement. 1–15 (2025) doi:10.1002/dad2.70074. Martino-Adami, P. V. et al. Prognostic value of Alzheimer’s disease plasma biomarkers in the oldest-old: a prospective primary care-based study. Lancet Reg. Heal. - Eur. 45 , 1–13 (2024). van Maurik, I. S. et al. Biomarker-based prognosis for people with mild cognitive impairment (ABIDE): a modelling study. Lancet Neurol. 18 , 1034–1044 (2019). Rubinov, M. & Sporns, O. Complex network measures of brain connectivity: Uses and interpretations. Neuroimage 52 , 1059–1069 (2010). Zhang, L. et al. Exploring Alzheimer’s disease: a comprehensive brain connectome-based survey. Psychoradiology 4 , (2024). Sanz-Leon, P., Knock, S. A., Spiegler, A. & Jirsa, V. K. Mathematical framework for large-scale brain network modeling in The Virtual Brain. Neuroimage 111 , 385–430 (2015). Deco, G. et al. How local excitation-inhibition ratio impacts the whole brain dynamics. J. Neurosci. 34 , 7886–7898 (2014). D’Angelo, E. D. & Jirsa, V. Neurosciences The quest for multiscale brain modeling. Trends Neurosci. xx , 1–14 (2022). Stefanovski, L. et al. Bridging Scales in Alzheimer’s Disease: Biological Framework for Brain Simulation With The Virtual Brain. Front. Neuroinform. 15 , 1–30 (2021). Jirsa, V. K. et al. The Virtual Epileptic Patient: Individualized whole-brain models of epilepsy spread. Neuroimage 145 , 377–388 (2017). Falcon, M. I. et al. Functional mechanisms of recovery after chronic stroke: Modeling with the virtual brain. eNeuro 3 , 202–208 (2016). Aerts, H. et al. Modeling brain dynamics in brain tumor patients using the virtual brain. eNeuro 5 , (2018). Marti-Juan, G. et al. Using The Virtual Brain to study the relationship between structural and functional connectivity in people with multiple sclerosis: a multicentre study. Mult. Scler. J. 28 , 262–264 (2022). Monteverdi, A. et al. Subject-specific features of excitation/inhibition profiles in neurodegenerative diseases. Front. Aging Neurosci. 14 , 1–17 (2022). Monteverdi, A. et al. Virtual brain simulations reveal network-specific parameters in neurodegenerative dementias. Front. Aging Neurosci. 15 , 1–15 (2023). Thomas Yeo, B. T. et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J. Neurophysiol. 106 , 1125–1165 (2011). Buckner, R. L., Krienen, F. M., Castellanos, A., Diaz, J. C. & Thomas Yeo, B. T. The organization of the human cerebellum estimated by intrinsic functional connectivity. J. Neurophysiol. 106 , 2322–2345 (2011). McKhann, G. the diagnosis of dementia due to Alzheimer’s disease. Alzheimers Dement 7 , 263–269 (2012). Capitani, E. & Laiacona, M. Composite neuropsychological batteries and demographic correction: Standardization based on equivalent scores, with a review of published data. J. Clin. Exp. Neuropsychol. 19 , 795–809 (1997). Nigri, A. et al. Quantitative MRI Harmonization to Maximize Clinical Impact: The RIN–Neuroimaging Network. Front. Neurol. 13 , (2022). Faust, T., Gunner, G. & Schafer, D. P. Mechanisms governing activity-dependent synaptic pruning in the mammalian CNS. Nat. Rev. Neurosci. 22 , 657–673 (2021). Berlot, R., Metzler-Baddeley, C., Ikram, M. A., Jones, D. K. & O’Sullivan, M. J. Global efficiency of structural networks mediates cognitive control in mild cognitive impairment. Front. Aging Neurosci. 8 , 1–11 (2016). Wong, K. F. & Wang, X. J. A recurrent network mechanism of time integration in perceptual decisions. J. Neurosci. 26 , 1314–1328 (2006). Deco, G., Jirsa, V. & Friston, K. J. The Dynamical and Structural Basis of Brain Activity. Principles of Brain Dynamics: Global State Interactions 0 at https://doi.org/10.7551/mitpress/9108.003.0003 (2012). Conca, F. et al. Clinical validity of the Italian adaptation of the Uniform Data Set Neuropsychological Test Battery (I-UDSNB) in Mild Cognitive Impairment and Alzheimer’s Disease. Alzheimer’s Res. Ther. 16 , 1–9 (2024). Chen, X., Toueg, T. N., Harrison, T. M., Baker, S. L. & Jagust, W. J. Regional tau deposition reflects different pathways of subsequent neurodegeneration and memory decline in cognitively normal older adults. Ann Neurol. 95 , 249–259 (2025). Kurth, S. et al. Alzheimer’s disease patients activate attention networks in a short-term memory task. NeuroImage Clin. 23 , 101892 (2019). Lorenzi, R. M. et al. A multi-layer mean-field model of the cerebellum embedding microstructure and population-specific dynamics. PLoS Comput. Biol. 19 , 1–30 (2023). Castellazzi, G. et al. A comprehensive assessment of resting state networks: Bidirectional modification of functional integrity in cerebro-cerebellar networks in dementia. Front. Neurosci. 8 , 1–18 (2014). Additional Declarations No competing interests reported. Supplementary Files supplementary.pdf Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6550081","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":452667659,"identity":"d7361678-7f5f-4b15-9005-529c11817398","order_by":0,"name":"Anita Monteverdi","email":"data:image/png;base64,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","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":true,"prefix":"","firstName":"Anita","middleName":"","lastName":"Monteverdi","suffix":""},{"id":452667660,"identity":"eca6fd38-61b8-4632-8360-1cd69b52d3e0","order_by":1,"name":"Matteo Cotta Ramusino","email":"","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":false,"prefix":"","firstName":"Matteo","middleName":"Cotta","lastName":"Ramusino","suffix":""},{"id":452667661,"identity":"630ca5b2-a54a-45f0-902a-3fc99a8c87f8","order_by":2,"name":"Francesca Conca","email":"","orcid":"","institution":"Institute for Advanced Studies (IUSS)","correspondingAuthor":false,"prefix":"","firstName":"Francesca","middleName":"","lastName":"Conca","suffix":""},{"id":452667662,"identity":"599bb454-60c1-4b3c-bf7b-71da106afe4d","order_by":3,"name":"Sofia Manzon","email":"","orcid":"","institution":"University of Pavia","correspondingAuthor":false,"prefix":"","firstName":"Sofia","middleName":"","lastName":"Manzon","suffix":""},{"id":452667664,"identity":"b8d46f10-16b0-49ea-964e-025a86a86e75","order_by":4,"name":"Alberto Redolfi","email":"","orcid":"","institution":"IRCCS Istituto Centro San Giovanni di Dio Fatebenefratelli","correspondingAuthor":false,"prefix":"","firstName":"Alberto","middleName":"","lastName":"Redolfi","suffix":""},{"id":452667666,"identity":"09c0a936-dbdf-4bf3-81ee-4cadce371079","order_by":5,"name":"Eleonora Lupi","email":"","orcid":"","institution":"University of Pavia","correspondingAuthor":false,"prefix":"","firstName":"Eleonora","middleName":"","lastName":"Lupi","suffix":""},{"id":452667668,"identity":"a1f99e49-67e3-4ffb-95cf-14d415dc8177","order_by":6,"name":"Marialaura De Grazia","email":"","orcid":"","institution":"University of Pavia","correspondingAuthor":false,"prefix":"","firstName":"Marialaura","middleName":"","lastName":"De Grazia","suffix":""},{"id":452667669,"identity":"ad3b7517-588c-45e8-81a0-370bee5dab81","order_by":7,"name":"Roberta Maria Lorenzi","email":"","orcid":"","institution":"University of Pavia","correspondingAuthor":false,"prefix":"","firstName":"Roberta","middleName":"Maria","lastName":"Lorenzi","suffix":""},{"id":452667670,"identity":"e397b3de-2c42-49b2-a6d1-847e90841afe","order_by":8,"name":"Marta Gaviraghi","email":"","orcid":"","institution":"University of Pavia","correspondingAuthor":false,"prefix":"","firstName":"Marta","middleName":"","lastName":"Gaviraghi","suffix":""},{"id":452667671,"identity":"21187d3d-a21c-4699-ae3e-1ff0d549dd8a","order_by":9,"name":"Laura Mazzocchi","email":"","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":false,"prefix":"","firstName":"Laura","middleName":"","lastName":"Mazzocchi","suffix":""},{"id":452667672,"identity":"a5ec429a-3528-427c-b68a-a088dccf64dc","order_by":10,"name":"Lisa M. Farina","email":"","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":false,"prefix":"","firstName":"Lisa","middleName":"M.","lastName":"Farina","suffix":""},{"id":452667674,"identity":"d7e05b8f-9e9d-46e8-8878-5871797907d1","order_by":11,"name":"Alfredo Costa","email":"","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":false,"prefix":"","firstName":"Alfredo","middleName":"","lastName":"Costa","suffix":""},{"id":452667676,"identity":"5ef6db50-58de-4a86-bf27-3698726ed529","order_by":12,"name":"Anna Pichiecchio","email":"","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":false,"prefix":"","firstName":"Anna","middleName":"","lastName":"Pichiecchio","suffix":""},{"id":452667677,"identity":"ce0d7063-e0f0-4838-b7fc-78146644ec1d","order_by":13,"name":"Stefano F. Cappa","email":"","orcid":"","institution":"Institute for Advanced Studies (IUSS)","correspondingAuthor":false,"prefix":"","firstName":"Stefano","middleName":"F.","lastName":"Cappa","suffix":""},{"id":452667678,"identity":"eac7bab5-7183-4bf8-beea-aadccf91e90c","order_by":14,"name":"Claudia A. M. Gandini Wheeler-Kingshott","email":"","orcid":"","institution":"NMR Research Unit, UCL Queen Square Institute of Neurology","correspondingAuthor":false,"prefix":"","firstName":"Claudia","middleName":"A. M. Gandini","lastName":"Wheeler-Kingshott","suffix":""},{"id":452667679,"identity":"c143993f-0bff-4379-80d3-cf668d593d20","order_by":15,"name":"Fulvia Palesi","email":"","orcid":"","institution":"University of Pavia","correspondingAuthor":false,"prefix":"","firstName":"Fulvia","middleName":"","lastName":"Palesi","suffix":""},{"id":452667680,"identity":"12f041c6-b993-4a6e-b4ff-003f552fd63f","order_by":16,"name":"Egidio D’Angelo","email":"","orcid":"","institution":"IRCCS Mondino Foundation","correspondingAuthor":false,"prefix":"","firstName":"Egidio","middleName":"","lastName":"D’Angelo","suffix":""}],"badges":[],"createdAt":"2025-04-28 17:53:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6550081/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6550081/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82582051,"identity":"2354a4cc-e033-4547-9e89-4f0037c3baf2","added_by":"auto","created_at":"2025-05-13 06:41:00","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":991554,"visible":true,"origin":"","legend":"\u003cp\u003eIntegrated TVB and GT analysis of resting-state networks.\u003c/p\u003e\n\u003cp\u003eThe figure reports the example of the Default Mode Network (DMN) in an MCI patient. Resting-state networks are extracted from DWI using tractography (structural connectome – SC) and rs-fMRI (functional and functional dynamics connectomes – FC and FCD). In the SC matrix, the regions are reported on the axes and the number of connecting tracts is color-coded (relative scale 0-0.45). In the FC matrices, the regions are reported on the axes and the PCC of the BOLD signals between pairs of regions is color-coded (relative scale 0-1). In the FCD matrices, time is reported on the axes and the PCC of the BOLD signals between regions is color-coded (relative scale 0-1). Explanatory templates for the SC and FC matrices are reported in Suppl. Fig.7. The structural and functional connectomes are then used to reconstruct brain dynamics in the virtual brain (TVB) and network topology using graph theory (GT). The avatar is made of nodes (mathematical representations of neuronal activity) and edges (connections between nodes). Connectivity strength and node parameters are optimized by comparison with the experimental FC and FCD, used as rs-fMRI template, during the model inversion process.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/a3e194268e1bc96c72b1d28b.png"},{"id":82580634,"identity":"8f3959fa-3ca6-432a-af09-b425679601f7","added_by":"auto","created_at":"2025-05-13 06:33:00","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":272212,"visible":true,"origin":"","legend":"\u003cp\u003eTopological and dynamical features at group level.\u003c/p\u003e\n\u003cp\u003eBoxplots of GT and TVB measures addressing \u003cstrong\u003e(A)\u003c/strong\u003e structural connectivity, \u003cstrong\u003e(B)\u003c/strong\u003efunctional connectivity, and \u003cstrong\u003e(C) \u003c/strong\u003edynamic parameters of DMN, LN, AN, VN in different groups (healthy controls=HC, mild cognitive impairment=MCI, Alzheimer’s disease=AD). Only networks showing a significant difference between groups (p\u0026lt;0.05, GLM Bonferroni corrected) are reported.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/437233ba111d7f30c5115e7d.png"},{"id":82580650,"identity":"167d3558-cb2c-4234-83f3-1971a221743b","added_by":"auto","created_at":"2025-05-13 06:33:01","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":123994,"visible":true,"origin":"","legend":"\u003cp\u003eClustering of topological features at single-subject level\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e Clusters (different colors) found by K-means using the meaningful topological features (GT measures of LN, AN and SMN) identified by RF. Bar plots show the frequency of each clinical group (HC, MCI, AD) and MCI+ and MCI- distributions in the clusters. Boxplots show global cognitive performance (assessed with MMSE) of MCI patients in the clusters. \u003cstrong\u003eB)\u003c/strong\u003eBoxplots of GT meaningful features in LN, AN and SMN in each cluster (same color-code as in A). Asterisks indicate significant differences (multivariate GLM, p\u0026lt;0.05) between clusters.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/266bc001a5ceddffc42e1bab.png"},{"id":82580633,"identity":"81a9896c-32fd-4d0b-876a-9564d5147d6f","added_by":"auto","created_at":"2025-05-13 06:33:00","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":94405,"visible":true,"origin":"","legend":"\u003cp\u003eClustering of dynamical features at single-subject level\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA) \u003c/strong\u003eClusters (different colors) found by K-means using the meaningful virtual brain derived features (TVB parameters of LN, DMN, AN, FPN, and SMN) identified by RF. Bar plots show the frequency of each clinical group (HC, MCI, AD) and MCI+ and MCI- distributions in the clusters. Boxplots show global cognitive performance (assessed with MMSE) of MCI patients in the clusters. \u003cstrong\u003eB) \u003c/strong\u003eBoxplots of TVB meaningful features in LN, DMN, AN, FPN and SMN in each cluster (same color-code as in A). Asterisks indicate significant differences (multivariate GLM, p\u0026lt;0.05) between clusters.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/b1fdb08d7a484f5011d5deb3.png"},{"id":82582052,"identity":"7723d48b-1908-46ed-8ead-ea0f390c26d1","added_by":"auto","created_at":"2025-05-13 06:41:00","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":208185,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation of topological and dynamical parameters with neuropsychological scores.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eA)\u003c/strong\u003e The bar plot shows the explained variance (% R\u003csup\u003e2\u003c/sup\u003e) in backward regressions of neuropsychological scores against structural, functional, or dynamic parameters in different brain networks (DMN, LN, AN, FPN, VN, SMN). All the reported values are statistically significant (F-test, p\u0026lt;0.05). It should be noted that GT and TVB parameters alone yielded a poorer correlation with neuropsychology than GT-TVB combined. The dashed line indicates a threshold at 50% of the explained variance. Note that this threshold is almost exclusively crossed by GT-TVB combined parameters. \u003cstrong\u003eB) \u003c/strong\u003eRegression plots for verbal working memory in the DMN (R\u003csup\u003e2\u003c/sup\u003e=0.6000, p\u0026lt;0.011), visual attention and task switching in the AN (R\u003csup\u003e2\u003c/sup\u003e=0.657, p\u0026lt;0.003), and semantic fluency in the LN (R\u003csup\u003e2\u003c/sup\u003e=0.6810, p\u0026lt;0.004), vs. combined GT-TVB scores.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/c21d136421bc7a45a05650fa.png"},{"id":84055731,"identity":"9b1cbe44-ff45-4c8a-abd2-a2b5d7277003","added_by":"auto","created_at":"2025-06-06 09:17:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2562297,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/21505efc-0315-4a4f-86d5-a7e55bb8371e.pdf"},{"id":82580631,"identity":"f5fa9a10-a3d1-4c19-bdb9-ff0eded34fd5","added_by":"auto","created_at":"2025-05-13 06:33:00","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1168035,"visible":true,"origin":"","legend":"","description":"","filename":"supplementary.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6550081/v1/74d4d221a90d77f38d9620d2.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Alterations in topological and dynamical parameters correlate with disease biomarkers and neuropsychological scores in prodromic stages of dementia","fulltext":[{"header":"Background","content":"\u003cp\u003eThe non-invasive investigation of changes in brain networks can contribute to our understanding of clinical conditions associated with high neurological, neuropsychological and neuropathological heterogeneity, such as mild cognitive impairment (MCI). The diagnosis of MCI on the basis of core clinical criteria\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e covers a range of widely different conditions and progression modalities. A recent meta-analysis based on 89 studies with a mean 5.2 year follow-up revealed a risk of progression to dementia of 41.5% in clinical and 27.0% in population-based studies, with dementia due to Alzheimer’s disease (AD) as the most common outcome\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. The MCI individuals remaining stable were 49.3% in clinical and 49.8% in population studies. Reversion to normal cognition was 8.7% in clinical and 28.2% in population studies. While biomarker positivity, including plasma-based measurements\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, increased the risk of progression to dementia, a considerable variation in speed of progression and in clinical phenotype was found also in MCI due to AD\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. While current evidence mostly addresses biomarkers of pathology and their correlation with brain structure (e.g. volume loss) and function (regional metabolism reduction due to synaptic loss), little is known about the large-scale organization of network dynamics.\u003c/p\u003e \u003cp\u003eThe structural and functional reorganization of brain network architecture with dementia progression has been assessed with graph theory (GT) measures\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e applied to resting state connectivity data, which increasingly demonstrate their sensitivity to topological alterations linked to cognitive decline\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. However, this type of analysis cannot address the reorganization of circuit dynamics. A critical issue that waits for a solution is that the group analysis strategy, frequently used to investigate brain pathologies, gets short in capturing the multifactorial nature and the heterogeneity of subject profiles typical of the dementia syndrome. Thus, a multiparametric analysis projected in a subject-specific space can be expected to provide better insight into the link between network structural and functional alterations and large-scale brain dynamics.\u003c/p\u003e \u003cp\u003eHere we have applied a recent technology to generate a virtual brain to simulate the neurophysiological underpinnings of network dynamics in a non-invasive and personalized way. The Virtual Brain (TVB)\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e is a data-driven modelling framework that yields information about the ensemble properties of cells and microcircuits starting from magnetic resonance imaging (MRI) data. TVB combines mesoscopic neural mass models of neural dynamics with structural MRI data to create a brain “avatar”, which is made of nodes (brain areas) and edges (axonal fiber tracts derived with tractography), to simulate large-scale brain dynamics. Thus, TVB links brain structure, function, and dynamics into a generative mechanistic model. Appropriate computational representations of local microcircuit functions enable TVB to quantify physiological parameters like NMDA and GABA receptor-mediated transmission and recurrent excitation\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e, which are especially important to address alterations of synaptic excitation and inhibition in neurodegeneration\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. The potential of TVB to capture pathological heterogeneity and foster personalized patient profiling has been demonstrated in epilepsy\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, stroke\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, brain tumors\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, multiple sclerosis\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e and advanced dementia stages\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, and is further extended here to the case of MCI.\u003c/p\u003e \u003cp\u003eThe capacity of TVB to simulate brain dynamics can be complemented by GT that can extract, from the same MRI data, topological information about network structural and functional organization. Here, we combined TVB with GT (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) to extract information from MRI data and project the parameters into a high-dimensional space. Networks supporting specific integrative and sensorimotor functions\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e [including the default mode network (DMN), limbic network (LN), attention network (AN), and somatomotor network (SMN)] were considered in a cohort including healthy controls (HC), MCI (both AD biomarker positive and negative, i.e., MCI\u003csup\u003e+\u003c/sup\u003e and MCI\u003csup\u003e−\u003c/sup\u003e) and AD. The aim was to investigate the potential of multiparametric analysis of MRI data to yield single subject profiles correlating network mechanisms with Aβ and τ, and neuropsychological scores.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e "},{"header":"Methods","content":"\u003cp\u003eSubjects\u003c/p\u003e\u003cp\u003eA total of 60 subjects were recruited at the IRCCS Mondino Foundation (Pavia, Italy): 18 HC (69 ± 5 years, 9 females), 22 MCI subjects (75 ± 6 years, 14 females), and 20 AD patients (71 ± 7 years, 16 females) (Suppl. Table\u0026nbsp;1). The informed consent was collected for all subjects and the study was approved by the local ethical committee and carried out in accordance with the Declaration of Helsinki. The diagnosis was based on current clinical criteria\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Etiological diagnosis of MCI was provided by cerebrospinal fluid (CSF) analysis (i.e. Aβ and τ protein levels) or amyloid PET: MCI\u003csup\u003e+\u003c/sup\u003e (10 subjects − 7A + T + by CSF analysis, 3A + by PET) and MCI\u003csup\u003e−\u003c/sup\u003e (12 subjects). The exclusion criteria were diagnosis of any neurological or psychiatric condition different from AD and secondary causes of cognitive decline such as metabolic, iatrogenic, toxic, and endocrine.\u003c/p\u003e\u003cp\u003eNeuropsychological assessment\u003c/p\u003e\u003cp\u003eAll subjects underwent a neuropsychological examination assessing the global cognitive function with the Mini-Mental State Examination (MMSE) and a standardized battery of tests to assess different cognitive functions. Neuropsychological tests assessed the following cognitive domains: verbal short-term memory (digit span forward), verbal working memory (digit span backward), verbal long-term memory (short story test and FCSRT), visuo-spatial long-term memory (Rey–Osterrieth complex figure recall), phonological and semantic fluency (Category and Phonemic Fluency tasks), visuo-constructional function (Rey-Osterrieth complex figure copy) and visual attention and task switching (trail making test part A and B). According to the reference norms for the Italian population, raw scores for each test were corrected for age, education and sex and converted into equivalent scores (ES) ranging from 0 to 4 (0 indicating a pathological performance)\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e (Suppl. Table\u0026nbsp;1).\u003c/p\u003e\u003cp\u003eStructural and functional connectivity\u003c/p\u003e\u003cp\u003eMRI acquisitions were performed using a 3T Siemens Skyra scanner with a 32-channel head coil. The protocol included diffusion weighted imaging (DWI) and resting-state fMRI (rs-fMRI) scans\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Preprocessing of DWI and rs-fMRI data were performed according to established techniques (see Supplementary Methods). To consider both cerebro-cortical, subcortical, and cerebellar regions, an ad-hoc gray matter parcellation atlas was created combining 93 AAL and 33 SUIT labels. The gray matter nodes were then remapped onto the Buckner (cerebellar) and Yeo (cerebral)\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e atlases identifying six networks known to support integrative functions (DMN, FPN, LN, AN) and sensorimotor functions (VN, SMN) relevant to MCI and AD. DWI and rs-fMRI data was then used to construct structural and functional connectivity (SC and FC) matrices of these networks (see Supplementary Methods).\u003c/p\u003e\u003cp\u003eTopological characterization of resting-state networks\u003c/p\u003e\u003cp\u003eFor each subject, the main GT metrics (Brain Connectivity Toolbox, Matlab) were obtained from static FC and SC matrices of resting state brain networks\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Parameters describing the arrangement of network nodes were the density and the core nodes. Structural and functional networks integration and segregation were evaluated with global efficiency and clustering coefficient values. Combining the path length and clustering coefficient measures allowed to define the small-worldness, which reflects the tendency of a network to balance local segregation with long-distance integration. As a measure of node centrality the betweenness centrality was computed from path lengths. For the SC matrix we further considered node strength, and shortest path length. For the FC matrix, we considered intraFC (see Supplementary Methods).\u003c/p\u003e\u003cp\u003eVirtual Brain modelling\u003c/p\u003e\u003cp\u003eFor each subject and each network, TVB simulations were performed using the SC matrix (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) and the nodes were endowed with Wong-Wang (WW) neural mass models\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e (Suppl. Figure\u0026nbsp;1). The WW is made of excitatory and inhibitory neural populations connected through NMDA synapses (J\u003csub\u003eNMDA\u003c/sub\u003e), and GABA-A synapses (J\u003csub\u003ei\u003c/sub\u003e), and recurrent excitation (w+). The SC matrix weights inter-node connections and is scaled by the global coupling (G), which denotes long-range coupling strength between nodes. All parameters were initially set as in Deco and colleagues\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e (Suppl. Table\u0026nbsp;2) and then tuned by model optimization. The simulated neural activity was convolved with the Ballon-Windkessel hemodynamic model to reconstruct the resting-state BOLD signal. Simulation generated an 8-minute rs-fMRI time-series per node to calculate the simulated FC and FCD matrices, which were updated iteratively (Suppl. Figure\u0026nbsp;2) against the experimental FC and FCD matrices. The metrics used for optimization were Pearson Correlation Coefficient (PCC) for FC and Kolmogorov-Smirnov distance (D\u003csub\u003eKS\u003c/sub\u003e) for FCD. An optimal simulation required the highest PCC and the lowest D\u003csub\u003eKS\u003c/sub\u003e minimizing the cost function \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:(1-PCC)+{D}_{KS}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eStatistics and machine learning\u003c/p\u003e\u003cp\u003eGT measures and TVB parameters for each network were tested for normality (Shapiro-Wilk, p \u0026lt; 0.05). Then, a multivariate general linear model (GLM) followed by post-hoc Bonferroni correction was applied to detect network topology and excitation/inhibition differences between groups correcting for age and gender differences. To reduce the parameter space a decision-tree-based Random Forest algorithm was applied to both GT and TVB extracting the parameters that best separated the clinical groups (AD, MCI, and HC). The top five uncorrelated features were used as input for the clustering analysis (Silhouette and the K-means tests, see Supplementary Methods). Differences between clusters were assessed using multivariate GLM. \u003cem\u003eA posteriori\u003c/em\u003e analysis of subjects’ distributions between clusters was used to determine the correspondence between mechanistic parameters (topology and dynamics), molecular biomarker positivity (Aβ and τ), and global cognitive score (MMSE). Multiple regression analysis was performed to explore the relationship between the neuropsychological scores of specific cognitive domains and GT/TVB parameters. The backward regression algorithm automatically removed predictors until selecting the significant ones (F test, p \u0026lt; 0.05) explaining the neuropsychological scores variance.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe subjects included in this study belonged to three groups: HC, MCI, and AD. All subjects underwent structural and functional MRI scans, which were used to reconstruct the structural and functional connectome and to perform GT measures and TVB simulations (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eStructural, functional and dynamical characterization of resting-state networks at different dementia stages (group comparison)\u003c/p\u003e \u003cp\u003eGT measures revealed the topological organization of resting-state networks (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). From a structural point of view (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA), MCI patients showed higher clustering coefficient, global efficiency, and nodal strength in the DMN compared to AD (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, Bonferroni corrected). From a functional point of view (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB), MCI patients showed higher values of clustering coefficient, global efficiency, and intra functional connectivity (intraFC) in the AN, VN and SMN compared to AD (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, Bonferroni corrected). AD patients also showed lower global efficiency, clustering coefficient and intraFC in AN, VN and SMN, accompanied by higher betweenness centrality in AN compared to HC.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTVB simulations yielded subject-specific parameters describing the dynamics of excitation and inhibition in resting-state networks. Model parameters related to excitation levels (J\u003csub\u003eNMDA\u003c/sub\u003e and w\u003csub\u003e+\u003c/sub\u003e) of the DMN showed significant differences between groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC), with AD patients showing higher J\u003csub\u003eNMDA\u003c/sub\u003e and lower w\u003csub\u003e+\u003c/sub\u003e compared to HC and MCI (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, Bonferroni corrected).\u003c/p\u003e \u003cp\u003eIn aggregate, at group level, none of the GT or TVB parameters showed significant differences between MCI and HC, while several differences emerged between HC or MCI and AD.\u003c/p\u003e \u003cp\u003eSubject-specific profiles of resting-state network topology and dynamics\u003c/p\u003e \u003cp\u003eParameters describing network topology and dynamics in GT and TVB analysis were given as input to machine learning algorithms (random forest for features selection followed by k-mean clustering and Silhouette analysis) to generate multiparametric subject-specific profiles. An \u003cem\u003ea posteriori\u003c/em\u003e analysis of subject-specific distributions between clusters was performed to evaluate the correspondence between GT and TVB parameters, Aβ and τ biomarkers (MCI\u003csup\u003e+\u003c/sup\u003e, MCI\u003csup\u003e\u0026minus;\u003c/sup\u003e), and cognitive decline (MMSE score) in MCI patients.\u003c/p\u003e \u003cp\u003eTopological characterization\u003c/p\u003e \u003cp\u003eTopological parameters of LN, AN and SMN (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) were identified by random forest as the most informative ones distinguishing pathological from healthy subjects (Suppl. Figure\u0026nbsp;3). The five most meaningful uncorrelated parameters were structural betweenness centrality of LN, node strength of LN, functional global efficiency of AN, structural global efficiency of SMN and functional small-worldness of LN. Based on these five parameters, Silhouette and K-means analysis identified three main topology-based clusters (Suppl. Figure\u0026nbsp;4). Ct1 was characterized by lower values of AN global efficiency (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and higher values of LN functional small-worldness compared to other clusters (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). This cluster mainly contained AD patients. Ct2 showed higher values of LN node strength (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and SMN structural global efficiency compared to the other clusters (p\u0026thinsp;\u0026lt;\u0026thinsp;0.005), while AN global efficiency was increased compared to Ct1 but decreased compared to Ct3 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This cluster mainly contained MCI patients which, in turn, were mostly belonging to MCI\u003csup\u003e+\u003c/sup\u003e (8 out of 9). Ct3 showed higher values of AN global efficiency (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and lower values of structural LN betweenness centrality compared to the other clusters (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). This cluster mainly contained HC subjects and MCI patients entirely belonging to MCI\u003csup\u003e\u0026minus;\u003c/sup\u003e (8 out of 8).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDynamic properties characterization\u003c/p\u003e \u003cp\u003eTVB parameters of LN, AN, DMN, FPN, and SMN (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) were identified by random forest as the most informative ones distinguishing pathological from healthy subjects (Suppl. Figure\u0026nbsp;5). The five most meaningful uncorrelated parameters were recurrent excitation (w\u003csub\u003e+\u003c/sub\u003e) of LN, AN, DMN, FPN, and global coupling (G) of SMN. Based on these five parameters, Silhouette and K-means analysis identified three clusters (Suppl. Figure\u0026nbsp;6). Cd1 was characterized by higher values of w\u003csub\u003e+\u003c/sub\u003e in LN (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and G in SMN compared to other clusters (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) and mainly contained HC and MCI patients. Cd2 showed lower values of w\u003csub\u003e+\u003c/sub\u003e in DMN compared to other clusters (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and contained most of AD patients (17 out of 21). Cd3 showed higher values of w\u003csub\u003e+\u003c/sub\u003e in the AN network compared to other clusters (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and mainly contained MCI patients. It should be noted that MMSE of MCI patients in Cd2 was significantly lower than MMSE of MCI patients in Cd1 or Cd3.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCorrespondence between multiparametric profiling, molecular biomarkers and cognitive decline\u003c/p\u003e \u003cp\u003eIt is worth recapitulating here the main correlations that emerged between parameters reported in Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. \u003cem\u003eA posteriori\u003c/em\u003e analysis of topology-based cluster composition showed that the majority of MCI\u003csup\u003e\u0026minus;\u003c/sup\u003e was in Ct3 together with the majority of HC, while MCI\u003csup\u003e+\u003c/sup\u003e mostly fell in Ct2, without a significant difference in MMSE score. The \u003cem\u003ea posteriori\u003c/em\u003e analysis of dynamics-based cluster composition showed heterogeneity of MCI\u003csup\u003e\u0026minus;\u003c/sup\u003e and MCI\u003csup\u003e+\u003c/sup\u003e between clusters. However, MCI patients belonging to Cd2 showed significantly lower MMSE compared to MCI patients in other clusters.\u003c/p\u003e \u003cp\u003eThe relationship of network topology and dynamics with neuropsychology\u003c/p\u003e \u003cp\u003eTo assess the correlation between GT and TVB parameters with subject-specific clinical severity, patient-specific neuropsychological scores were used in a backward regression model against network-specific parameters. Interestingly, combinations of GT and TVB parameters were more effective than individual measures in explaining the variation of neuropsychological scores across multiple cognitive domains (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA), with an explained variance (adjusted R\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e) reaching in some cases\u0026thinsp;~\u0026thinsp;70%. Examples of these correlations for visual attention and task switching in the AN, verbal working memory in the DMN, and semantic fluency in the LN, are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis work provides the first evidence that a subject-specific profiling of MCI patients can be performed through the combination of GT, which describes the topological organization of networks, and TVB, which simulates network dynamics. This approach identifies a rich set of network alterations that allows to identify clusters of MCI patients positive to AD molecular biomarkers (Aβ and τ) or sensitive to global cognitive changes (MMSE), and to correlate microscopic network parameters with neuropsychological scores. Conversely, no differences between MCI and HC did emerge at the group level (not unexpectedly, differences emerged between HC or MCI and AD, though). Thus, the GT/TVB multiparametric analysis extending over multiple network properties and cognitive domains in single patients is superior to classical low-dimensional analysis at group level for addressing the well-known heterogeneity of the clinically defined MCI condition.\u003c/p\u003e \u003cp\u003eThe topological analysis (GT) revealed a reorganization of LN, AN, and SMN. A crucial finding is that subjects were grouped into three clusters, one hosting most MCI\u003csup\u003e+\u003c/sup\u003e and another most MCI\u003csup\u003e\u0026minus;\u003c/sup\u003e patients, suggesting that, even if clinical severity is not different, molecular biomarker positivity is already assigning some of the MCI subjects to the prodromal phase of AD. The AN functional global efficiency decreased with the AD/HC ratio in the clusters suggesting that a decreased inter-regional communication plays a role in the cognitive decline from HC to MCI to AD. The LN functional small-worldness increased with the AD/HC ratio in the clusters suggesting a desynchronization of distant brain regions associated with an increase of short-distance synchronization. The LN higher betweenness centrality in the clusters with high AD/HC ratio may facilitate network integration possibly implying mechanisms compensating for the detrimental effect of small-worldness. The LN high node strength in the cluster with the majority of AD and MCI\u003csup\u003e+\u003c/sup\u003e patients may also play a compensatory role against synaptic pruning\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. The SMN showed the highest structural global efficiency in the cluster containing most of MCI\u003csup\u003e+\u003c/sup\u003e suggesting that it might enhance cognitive control\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e compensating the derangement in AN and LN. In aggregate, specific changes in AN, LN, and SMN unveil a profound alteration of the integration/segregation balance in MCI along with possible compensation antagonizing the evolution to dementia.\u003c/p\u003e \u003cp\u003eThe dynamical analysis (TVB) revealed marked changes in the recurrent excitation of LN, AN, and DMN, and in the global coupling of SMN, allowing to group subjects into three clusters (distinct from those obtained with topological analysis). Here, while the segregation of MCI\u003csup\u003e+\u003c/sup\u003e and MCI\u003csup\u003e\u0026minus;\u003c/sup\u003e is less clear-cut, there is a significant reduction of the MMSE score in MCI subjects belonging to the cluster with the highest AD/HC ratio and the lowest values of network recurrent excitation. Consistently, an extended recurrent excitation in cortical circuits boosts local processing supporting decision making and working memory tasks\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e, which are reduced in MCI and even more in AD. The higher G value in the SMN implies an increased network synchronization\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, a sign of synaptic dysfunction already reported in AD\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e that is now extended to MCI patients. It should also be noted that FPN, which is involved in frontotemporal dementia but not in AD, did not show significant changes of recurrent excitation, in agreement with previous studies\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe combination of topological and dynamical parameters (GT and TVB) explained neuropsychological scores variance (up to 70%) in multiple cognitive domains better than the individual measures alone, implying that the measures are (at least partly) independent and provide an incremental amount of information. Consistently, visual attention and task switching, semantic fluency, and verbal short-term memory correlated with parameter changes in DMN, AN, and LN, confirming a main involvement of these networks in the worsening of cognitive functions\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn aggregate, while classical group analysis captures changes in the AD brain compared to MCI and HC, single subject multiparametric analysis reclassifies subjects, unveiling changes in MCI that correlate with pathology biomarkers (GT) and the global cognitive state (TVB). Not unexpectedly, given the different sensitivity of the two techniques, pathology markers reflect tissue damage while brain dynamics reflect cognitive processing. The low integration/segregation balance and recurrent excitation in MCI\u003csup\u003e+\u003c/sup\u003e and AD patients places DMN, AN, and LN networks, which are responsible for memory and executive functions, at the core of the pathogenetic process. Tau PET imaging studies show that the extension of tau pathology from the temporo-mesial region towards neocortical temporal and parietal regions is a strong predictor of progression to dementia\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. It is therefore not surprising that DMN, LN and AN, which include temporo-mesial and temporo-parietal areas, could be among the first and most affected networks in the clusters of patients with the majority of MCI\u003csup\u003e+\u003c/sup\u003e (7/10) and AD (17/20). The SMN, at odd with cognitive decline in MCI, shows enhanced global efficiency and may therefore be involved in compensatory processes, as also previously suggested for the AN\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThis study should be seen as the beginning of a broader set of investigations and can be improved in several respects. First, although the sample size does not impact \u003cem\u003eper se\u003c/em\u003e on the TVB ability of uncovering excitation/inhibition profiles, the study of a larger cohort is warranted. Moreover, GT and TVB analysis may be refined by curating tractography and using specific circuit models\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e, which may impact on structural and functional connectivity and brain dynamics. Specific aspect that may be considered further concern the role of the cerebellum, whose functional connectivity is markedly increased in MCI\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e, and of AN, which may better divided into the dorsal and ventral attention networks (DAN and VAN).\u003c/p\u003e \u003cp\u003eIn conclusion, the combined GT/TVB analysis allows to detect a set of changes in the DMN, LN, AN, and SMN of MCI patients presumably reflecting a combination of pathogenetic alterations and compensatory mechanisms. Then, multiparametric profiling identifies clusters of subjects with high expression of molecular biomarkers and reduced MMSE scores. In a clinical perspective, a longitudinal analysis of GT/TVB results is warranted to predict who, among the MCI patients, will evolve into AD and benefit from a timely intervention, opening new perspectives for personalized treatments in prodromal dementia stages.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003eData sharing statement\u003c/p\u003e\n\u003cp\u003eAll codes used for brain dynamics simulations with TheVirtualBrain are available as a Python code that can be found at https://www.thevirtualbrain.org/tvb/zwei. The dataset will be made available on Zenodo.\u003c/p\u003e\n\u003cp\u003eDeclaration of interest\u003c/p\u003e\n\u003cp\u003eAuthors report no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Acknowledgements\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis work was performed at the IRCCS Mondino Foundation and was supported by the Italian Ministry of Health (SG-2021-12374430) to AM. ED\u0026rsquo;A acknowledges #NEXTGENERATIONEU (NGEU) and the Ministry of University and Research (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006) \u0026ndash; A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022) and EBRAINS-Italy (Project IR0000011, CUP B51E22000150006). CW-K acknowledges BRC (#BRC704/CAP/CGW), MRC (#MR/S026088/1), Ataxia UK, Rosetrees Trust (#PGL22/100041 and #PGL21/10079). CGWK is a shareholder in Queen Square Analytics Ltd. EL is a PhD student enrolled in the National PhD in Artificial Intelligence, XXXVIII cycle, course on Health and life sciences, organized by Universit\u0026agrave; Campus Bio-Medico di Roma. MG acknowledges \u0026ldquo;National Centre for HPC, Big Data and Quantum Computing\u0026rdquo; (Project CN00000013 PNRR MUR - M4C2 - Fund 1.4 - Call \u0026ldquo;National Centers\u0026rdquo; - law decree n. 3138 16 December 2021).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Funding\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Italian Ministry of Health (SG-2021-12374430).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAlbert S. Marylin, D. T. S. et al. The diagnosis of mild cognitive impairment due to Alzheimer\u0026rsquo;s disease: Recommendations from the National Institute on Aging- Alzheimer\u0026rsquo;s Association workgroups on diagnostic guidelines for Alzheimer\u0026rsquo;s disease. \u003cem\u003eAlzheimers Dement. 2011 May ; 7(3) 270\u0026ndash;279. doi10.1016/j.jalz.2011.03.008.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, (2011).\u003c/li\u003e\n\u003cli\u003eSalemme S, Lombardo FL, Lacorte E, et al. The prognosis of mild cognitive impairment : A systematic review and meta-analysis. \u003cem\u003eAlzheimer\u0026rsquo;s Dement.\u003c/em\u003e 1\u0026ndash;15 (2025) doi:10.1002/dad2.70074.\u003c/li\u003e\n\u003cli\u003eMartino-Adami, P. V. \u003cem\u003eet al.\u003c/em\u003e Prognostic value of Alzheimer\u0026rsquo;s disease plasma biomarkers in the oldest-old: a prospective primary care-based study. \u003cem\u003eLancet Reg. Heal. - Eur.\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 1\u0026ndash;13 (2024).\u003c/li\u003e\n\u003cli\u003evan Maurik, I. S. \u003cem\u003eet al.\u003c/em\u003e Biomarker-based prognosis for people with mild cognitive impairment (ABIDE): a modelling study. \u003cem\u003eLancet Neurol.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1034\u0026ndash;1044 (2019).\u003c/li\u003e\n\u003cli\u003eRubinov, M. \u0026amp; Sporns, O. Complex network measures of brain connectivity: Uses and interpretations. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cstrong\u003e52\u003c/strong\u003e, 1059\u0026ndash;1069 (2010).\u003c/li\u003e\n\u003cli\u003eZhang, L. \u003cem\u003eet al.\u003c/em\u003e Exploring Alzheimer\u0026rsquo;s disease: a comprehensive brain connectome-based survey. \u003cem\u003ePsychoradiology\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eSanz-Leon, P., Knock, S. A., Spiegler, A. \u0026amp; Jirsa, V. K. Mathematical framework for large-scale brain network modeling in The Virtual Brain. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cstrong\u003e111\u003c/strong\u003e, 385\u0026ndash;430 (2015).\u003c/li\u003e\n\u003cli\u003eDeco, G. \u003cem\u003eet al.\u003c/em\u003e How local excitation-inhibition ratio impacts the whole brain dynamics. \u003cem\u003eJ. Neurosci.\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 7886\u0026ndash;7898 (2014).\u003c/li\u003e\n\u003cli\u003eD\u0026rsquo;Angelo, E. D. \u0026amp; Jirsa, V. Neurosciences The quest for multiscale brain modeling. \u003cem\u003eTrends Neurosci.\u003c/em\u003e \u003cstrong\u003exx\u003c/strong\u003e, 1\u0026ndash;14 (2022).\u003c/li\u003e\n\u003cli\u003eStefanovski, L. \u003cem\u003eet al.\u003c/em\u003e Bridging Scales in Alzheimer\u0026rsquo;s Disease: Biological Framework for Brain Simulation With The Virtual Brain. \u003cem\u003eFront. Neuroinform.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1\u0026ndash;30 (2021).\u003c/li\u003e\n\u003cli\u003eJirsa, V. K. \u003cem\u003eet al.\u003c/em\u003e The Virtual Epileptic Patient: Individualized whole-brain models of epilepsy spread. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cstrong\u003e145\u003c/strong\u003e, 377\u0026ndash;388 (2017).\u003c/li\u003e\n\u003cli\u003eFalcon, M. I. \u003cem\u003eet al.\u003c/em\u003e Functional mechanisms of recovery after chronic stroke: Modeling with the virtual brain. \u003cem\u003eeNeuro\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 202\u0026ndash;208 (2016).\u003c/li\u003e\n\u003cli\u003eAerts, H. \u003cem\u003eet al.\u003c/em\u003e Modeling brain dynamics in brain tumor patients using the virtual brain. \u003cem\u003eeNeuro\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eMarti-Juan, G. \u003cem\u003eet al.\u003c/em\u003e Using The Virtual Brain to study the relationship between structural and functional connectivity in people with multiple sclerosis: a multicentre study. \u003cem\u003eMult. Scler. J.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 262\u0026ndash;264 (2022).\u003c/li\u003e\n\u003cli\u003eMonteverdi, A. \u003cem\u003eet al.\u003c/em\u003e Subject-specific features of excitation/inhibition profiles in neurodegenerative diseases. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1\u0026ndash;17 (2022).\u003c/li\u003e\n\u003cli\u003eMonteverdi, A. \u003cem\u003eet al.\u003c/em\u003e Virtual brain simulations reveal network-specific parameters in neurodegenerative dementias. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1\u0026ndash;15 (2023).\u003c/li\u003e\n\u003cli\u003eThomas Yeo, B. T. \u003cem\u003eet al.\u003c/em\u003e The organization of the human cerebral cortex estimated by intrinsic functional connectivity. \u003cem\u003eJ. Neurophysiol.\u003c/em\u003e \u003cstrong\u003e106\u003c/strong\u003e, 1125\u0026ndash;1165 (2011).\u003c/li\u003e\n\u003cli\u003eBuckner, R. L., Krienen, F. M., Castellanos, A., Diaz, J. C. \u0026amp; Thomas Yeo, B. T. The organization of the human cerebellum estimated by intrinsic functional connectivity. \u003cem\u003eJ. Neurophysiol.\u003c/em\u003e \u003cstrong\u003e106\u003c/strong\u003e, 2322\u0026ndash;2345 (2011).\u003c/li\u003e\n\u003cli\u003eMcKhann, G. the diagnosis of dementia due to Alzheimer\u0026rsquo;s disease. \u003cem\u003eAlzheimers Dement\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 263\u0026ndash;269 (2012).\u003c/li\u003e\n\u003cli\u003eCapitani, E. \u0026amp; Laiacona, M. Composite neuropsychological batteries and demographic correction: Standardization based on equivalent scores, with a review of published data. \u003cem\u003eJ. Clin. Exp. Neuropsychol.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 795\u0026ndash;809 (1997).\u003c/li\u003e\n\u003cli\u003eNigri, A. \u003cem\u003eet al.\u003c/em\u003e Quantitative MRI Harmonization to Maximize Clinical Impact: The RIN\u0026ndash;Neuroimaging Network. \u003cem\u003eFront. Neurol.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eFaust, T., Gunner, G. \u0026amp; Schafer, D. P. Mechanisms governing activity-dependent synaptic pruning in the mammalian CNS. \u003cem\u003eNat. Rev. Neurosci.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 657\u0026ndash;673 (2021).\u003c/li\u003e\n\u003cli\u003eBerlot, R., Metzler-Baddeley, C., Ikram, M. A., Jones, D. K. \u0026amp; O\u0026rsquo;Sullivan, M. J. Global efficiency of structural networks mediates cognitive control in mild cognitive impairment. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1\u0026ndash;11 (2016).\u003c/li\u003e\n\u003cli\u003eWong, K. F. \u0026amp; Wang, X. J. A recurrent network mechanism of time integration in perceptual decisions. \u003cem\u003eJ. Neurosci.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1314\u0026ndash;1328 (2006).\u003c/li\u003e\n\u003cli\u003eDeco, G., Jirsa, V. \u0026amp; Friston, K. J. The Dynamical and Structural Basis of Brain Activity. \u003cem\u003ePrinciples of Brain Dynamics: Global State Interactions\u003c/em\u003e 0 at https://doi.org/10.7551/mitpress/9108.003.0003 (2012).\u003c/li\u003e\n\u003cli\u003eConca, F. \u003cem\u003eet al.\u003c/em\u003e Clinical validity of the Italian adaptation of the Uniform Data Set Neuropsychological Test Battery (I-UDSNB) in Mild Cognitive Impairment and Alzheimer\u0026rsquo;s Disease. \u003cem\u003eAlzheimer\u0026rsquo;s Res. Ther.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 1\u0026ndash;9 (2024).\u003c/li\u003e\n\u003cli\u003eChen, X., Toueg, T. N., Harrison, T. M., Baker, S. L. \u0026amp; Jagust, W. J. Regional tau deposition reflects different pathways of subsequent neurodegeneration and memory decline in cognitively normal older adults. \u003cem\u003eAnn Neurol.\u003c/em\u003e \u003cstrong\u003e95\u003c/strong\u003e, 249\u0026ndash;259 (2025).\u003c/li\u003e\n\u003cli\u003eKurth, S. \u003cem\u003eet al.\u003c/em\u003e Alzheimer\u0026rsquo;s disease patients activate attention networks in a short-term memory task. \u003cem\u003eNeuroImage Clin.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 101892 (2019).\u003c/li\u003e\n\u003cli\u003eLorenzi, R. M. \u003cem\u003eet al.\u003c/em\u003e A multi-layer mean-field model of the cerebellum embedding microstructure and population-specific dynamics. \u003cem\u003ePLoS Comput. Biol.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1\u0026ndash;30 (2023).\u003c/li\u003e\n\u003cli\u003eCastellazzi, G. \u003cem\u003eet al.\u003c/em\u003e A comprehensive assessment of resting state networks: Bidirectional modification of functional integrity in cerebro-cerebellar networks in dementia. \u003cem\u003eFront. Neurosci.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1\u0026ndash;18 (2014).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"brain dynamics, excitatory/Inhibitory balance, mild cognitive impairment, Alzheimer’s disease, resting-state networks, virtual brain modelling, graph theory","lastPublishedDoi":"10.21203/rs.3.rs-6550081/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6550081/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBACKGROUND\u003c/h2\u003e \u003cp\u003eMild cognitive impairment (MCI) is a clinical condition in the continuum between normal cognition and dementia. Despite numerous studies, the heterogeneity of the underlying pathophysiology prevents a precise prediction of clinical evolution.\u003c/p\u003e\u003ch2\u003eMETHODS\u003c/h2\u003e \u003cp\u003eIn a cohort composed of MCI, healthy controls (HC), and Alzheimer\u0026rsquo;s disease (AD) patients, graph theory (GT) was combined with virtual brain modelling (TVB) to extract the information on network topology and dynamics embedded in magnetic resonance imaging (MRI) data. With this approach, the analysis was extended to a multiparametric space and brought from the group to the individual subject level.\u003c/p\u003e\u003ch2\u003eRESULTS\u003c/h2\u003e \u003cp\u003eThe comparison of network properties in HC, MCI, and AD revealed a profound reshaping of brain connectivity, which mainly affected the default mode, limbic, attention, and somatosensory networks. Interestingly, positivity to AD biomarkers (Aβ and τ) in MCI correlated with network topology, while a TVB parameter (i.e., recurrent excitation) correlated with reduced global cognition (MMSE score). There was a high correlation (R\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e\u0026thinsp;~\u0026thinsp;70%) between GT and TVB parameters and neuropsychological performance in multiple cognitive domains.\u003c/p\u003e\u003ch2\u003eCONCLUSIONS\u003c/h2\u003e \u003cp\u003eThe combination of GT and TVB parameters was superior to the individual techniques alone in providing a subject-specific phenotype of MCI sensitive to molecular biomarkers and correlated with neuropsychological scores. This, in turn, could form the basis for a more precise MCI stratification leading, in the future, to a personalized prediction of evolution and therapeutic intervention.\u003c/p\u003e","manuscriptTitle":"Alterations in topological and dynamical parameters correlate with disease biomarkers and neuropsychological scores in prodromic stages of dementia","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-13 06:32:55","doi":"10.21203/rs.3.rs-6550081/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a27abac7-e2a7-4936-85d0-e321e75d6ce7","owner":[],"postedDate":"May 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-08T15:38:29+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-13 06:32:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6550081","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6550081","identity":"rs-6550081","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}