Using hemodynamic parameters quantified with computational fluid 1 dynamics to explore potential mechanisms of occurrence and progression 2 in Alzheimer Disease

Background: It has gradually recognized that the patients with Alzheimer’s disease 15 (AD) have cerebral hemodynamic disorders. The purpose of the present study was to 16 exploit a novel computational fluid dynamics (CFD) model, which could be used to 17 measure intracranial hemodynamics quantitively in AD patients and to further explore 18 how the hemodynamic changes are involved in progression of AD. 19 Methods : A novel CFD model was constructed by personal magnetic resonance 20 angiography (MRA), vessel ultrasound and blood pressure value of all subjects, of 21 whom included AD patients, vascular dementia (VaD) patients and well-matched healthy controls (HCs). Demographic, clinical and imaging data of all subjects were 23 recorded and analyzed. Quantitative total cerebral blood flow (CBF) and 24 cerebrovascular resistance (CVR) were compared among three groups, in order to 25 ascertain the potential hemodynamic disorders in AD patients. 26 Results: Total CBF and CVR of AD patients were significantly different from those of 27 HCs (both P <0.01), but not different from patients with VaD (both P >0.5), despite the 28 cerebral arteries in AD patients were anatomically intact. Total CBF was negatively 29 correlated with total CVR (r s =-0.822, P <0.001) in AD patients. Comparing with HCs, 30 Elevated CVR ( OR =2.25, P =0.004) and age ( OR =2.06, P =0.021) were independent 31 risk factor of AD. 32 Conclusions: CFD can be applied to non-invasively and conveniently quantify and 33 visualize biomechanical changes of cerebral blood flow. Patients with AD have 34 dysfunction of cerebral hemodynamic, including lower CBF and higher CVR, and the 35 CVR was an independent risk factor of AD. These findings provide quantitative 36 evidence to support that increase of cerebrovascular resistance may involve in 37 development of AD.

healthy controls (HCs). Demographic, clinical and imaging data of all subjects were 23 recorded and analyzed. Quantitative total cerebral blood flow (CBF) and 24 cerebrovascular resistance (CVR) were compared among three groups, in order to 25 ascertain the potential hemodynamic disorders in AD patients.  Conclusions: CFD can be applied to non-invasively and conveniently quantify and 33 visualize biomechanical changes of cerebral blood flow. Patients with AD have 34 dysfunction of cerebral hemodynamic, including lower CBF and higher CVR, and the 35 CVR was an independent risk factor of AD. These findings provide quantitative clinicians, lifestyle and vascular risk factors accelerate VaD progression [4] . However, 48 recently studies indicated that cardiovascular risk factors correlate with the occurrence 49 and development of AD [5] . For example, previous studies have confirmed higher 50 vascular risk and lower physical activity are associated with burden of β-Amyloid and 51 cognitive decline [6,7] . 52 The circulatory pathophysiological changes mediated by vascular risk factors were 53 always accompanied by intracranial hemodynamic disorder, which was involved in 54 mechanisms of cognitive decline [5][6][7] . For instance, patients with cardiac dysfunction 55 manifest hemodynamic disorders and decreased cerebral perfusion, which subsequently 56 lead to injury or death of neurons [8,9] . Moreover, remodeling and cerebral vasomotor 57 disorders of intracranial or extracranial vessels reduce cerebral perfusion and increase 58 resistance of cerebral arteries, which impair metabolism of nervous tissue and clearance 59 of A-β amyloid, further exacerbate cognitive decline [10][11][12] . Therefore detection of 60 hemodynamic disorders may contribute to identification potentially pathophysiological 61 changes in dementia patients.

62
Hemodynamic parameters can be measured indirectly through some medical patients [13] , AD patients also exhibit increased CVR index (CVRi) and diminished CBF 68 in inferior parietal and temporal cerebral [14,15] . However, ASL has poor noise to signal 69 ratio and only reflects changes in a small portion of hemodynamic parameters. Flow 70 velocity and the pulsatility index can be evaluated by TCD, in which, increased CVRi 71 were found in aged adults [16] and AD patients [17] . A meta-analysis indicated that 72 hemodynamic disturbance in VaD was more severe than that of AD [18] . However, TCD 73 cannot accurately detect hemodynamic parameters of distal arterial branches, 74 furthermore, the accuracy of TCD relies on an experienced operator and interpreter.

75
PET only measures the CBF and is applied limitedly. The 4D flow MRI is an emerging 76 imaging paradigm and capable to quantify the temporal evolution of complex blood 77 flow patterns within an acquired 3D volume, by which AD patients have been found to 78 have decreased mean flow in the internal carotid and middle cerebral arteries [19] .

79
However, there is a trade-off between the spatial and temporal resolution of 4D-flow 80 MRI, it is suitable either for the large arteries with fast velocity or the narrow vessels 81 with slow velocity, such as measurements of blood flow velocity in the aorta or veins, 82 thereby limiting the applications of 4D-flow MRI in cerebral arteries.

83
CFD is a well-established technique that provides comprehensive information of 84 hemodynamics non-invasively. Various 3D CFD models using routinely available 85 medical imaging had been proposed and applied to evaluate hemodynamic parameters, 86 for example, fractional flow reserve (FFR) was calculated by CFD based on computed 87 tomography angiography (CTA) , which has been approved to assess the risk of 88 coronary stenosis, and CFD derived FFR is highly comparable with the FFR measured 89 by a interventional pressure wire [20,21] . CFD technique can reduce unnecessary 90 interventional angiography effectively and help doctors to diagnose pathological 91 conditions [22,23] . Moreover, CFD can be applied to assess the risk of rupture and   Diseases-10 (ICD-10). Individuals who were cognitively normal were also included to 107 be HCs. All participants received MRI+MRA and ultrasound of cervical arteries.

108
Subjects were excluded from the study if they suffered from heavy organ dysfunction, 109 or a history of cognitive disorders. The study was approved by the ethics committee of  (Table 1 and Supplementary Table S1). The 10-year risk of 115 heart disease or stroke was determined using the ASCVD algorithm (website:    Modelling of blood flow in 3D 146 The blood flow was assumed to be a viscous and incompressible Newtonian fluid, 147 the heat transfer and compressibility effects of the vascular wall were neglected in this 148 process. The blood flow were defined as a constant density = 1.06 × 10 ! . "# 149 and dynamic viscosity = 3.5 × 10 "! . "# . "# , as the simulated blood flow was 150 not sensitive to these parameters [25,26] . A typical carotid artery diameter = 151 6.0 × 10 "! and its corresponding velocity of blood flow = 0.4 . "# were 152 assumed in order to calculate the Reynolds number: = / ≈ 121 , which 153 suggested that the blood flow was laminar. A 3D unsteady incompressible Navier- Stokes equation was then utilized to describe the blood flow, as follows: The equation for conservation of mass was defined as: where was the velocity vector, was the pressure, and was force of the body, 159 assumed equal to 0. where 5. represents the cross-sectional area of the artery at the inlet, as the 176 hemodynamic assumption resulted in a Poiseuille velocity profile, which is parabolic [27] .

177
The cross-sectional area was calculated by 5. = • (  hemodynamic parameters of all subjects were calculated by the 3D CFD model (Table   231 2). As compared with HCs, there were significant reduced total CBF or increased total displayed in the first and the second row, respectively. The first column ( fig a and b )is for an AD 238 patients, the second column (fig c and d )is for a VaD patients, and the third column (fig e and f) is 239 for a healthy subject. It is evident that the AD patient and the healthy subject are with intact arterial 240 trees, whereas VaD the patient is with scarce arterial branches. However, according to computation, 12 the total blood flow in the models was 692 ml/min (AD patient), 647 ml/min (VaD patient), and 998 242 ml/min (healthy subject) respectively. 243 Interactive associations of the hemodynamic parameters and risk factors 244 Bivariate Spearman correlation showed that total CBF was negatively correlated 245 with total CVR in whole subjects (fig 2a, rs=-0.826, P<0.001) and AD groups(fig 2b,   246 rs=-0.822, P<0.001). There were significant correlations between age and total CBF 247 (fig 2c, rs=-0.282, P<0.05) or total CVR(fig 2d, rs=0.278, P<0.05 259 Binary regression demonstrated that age (10-year increment; P=0.021) and CVR 260 (P=0.004) were independent risk factors for AD (Table 3). Independent risk factors of  During undertaking cognitive task, healthy subjects and stroke patients exhibited 281 a significant increase both in CBF and blood stream velocity [10,32] , which suggested 282 that cerebrovascular circulation adjusts its hemodynamic response to metabolic 283 requirements. However, the total CBF of the internal carotid and vertebral arteries were 284 decreased in VaD patients [28] , Furthermore, a marked decreased CBF in the parietal and 285 14 frontal cortex of AD or VaD patients has been observed, which was associated with 286 increased subcortical white matter lesions in VaD patients [33] . Stabilized CBF is 287 dependent on heart function and resistance of intracranial vessels[9], the CVRi of 288 middle cerebral arteries [17] , cortex and subcortex were increased in AD patients,  However previous studies only analyzed the correlation between AD and CVR or 294 CBF respectively. In present study, the decreased total CBF and increased CVR were 295 observed in AD group, and total CVR was an independent risk factor of AD, more 296 importantly, the total CBF was significantly and negatively correlated with total CVR.

297
Therefore, the CBF may be regulated by CVR. All above results demonstrated that the 298 increase of vascular resistance may affect the perfusion of whole brain and occurrence 299 of AD. Therefore, early discovery of changes in CVR indicates that potential 300 cerebrovascular lesions in AD.

301
The increases of cerebral resistance in AD patients are caused by other potential 302 mechanisms. Recent research confirmed that capillary constriction caused by Aβ 303 induces energy lack and neurodegeneration in neuron [35] , which subsequently elevate 304 the cerebral vascular resistance. Moreover the cerebral vascular resistance may be 305 increased by mixed brain lesions and remodeling of cerebral microvasculature which 306 were mediated by vascular risk factors [6,7] . Consequently the treatments of AD should There are some limitations to this study. Firstly, it is a cross-sectional research 312 study, the correlation between hemodynamic parameters and AD need to be verified by 313 follow-up studies. In a future study we will verify the correlation between more     We wish to thank all participants in this study, without whose agreement this research    The funding bodies had no role in the design of the study, data collection, ana-lysis, 358 interpretation, or writing of the manuscript.

359
Availability of data and materials 360 The dataset used during the current study is available from the corresponding author on 361 reasonable request.

362
Ethics approval and consent to participate 363 The study was approved by the Institutional Ethical Committee of Nanjing ZhongDa 364 Hospital, Southeast University and the participants gave written informed consents 365 prior to obtain the data. 367 Not applicable 368