Collaborative Uncertainty Benefits Multi-Agent Multi-Modal Trajectory Forecasting.

In multi-modal multi-agent trajectory forecasting, two major challenges have not been fully tackled: 1) how to measure the uncertainty brought by the interaction module that causes correlations among the predicted trajectories of multiple agents; 2) how to rank the multiple predictions and select the optimal predicted trajectory. In order to handle the aforementioned challenges, this work first proposes a novel concept, collaborative uncertainty (CU), which models the uncertainty resulting from interaction modules. Then we build a general CU-aware regression framework with an original permutation-equivariant uncertainty estimator to do both tasks of regression and uncertainty estimation. Furthermore, we apply the proposed framework to current SOTA multi-agent multi-modal forecasting systems as a plugin module, which enables the SOTA systems to: 1) estimate the uncertainty in the multi-agent multi-modal trajectory forecasting task; 2) rank the multiple predictions and select the optimal one based on the estimated uncertainty. We conduct extensive experiments on a synthetic dataset and two public large-scale multi-agent trajectory forecasting benchmarks. Experiments show that: 1) on the synthetic dataset, the CU-aware regression framework allows the model to appropriately approximate the ground-truth Laplace distribution; 2) on the multi-agent trajectory forecasting benchmarks, the CU-aware regression framework steadily helps SOTA systems improve their performances. Especially, the proposed framework helps VectorNet improve by 262 cm regarding the Final Displacement Error of the chosen optimal prediction on the nuScenes dataset; 3) in multi-agent multi-modal trajectory forecasting, prediction uncertainty is proportional to future stochasticity; 4) the estimated CU values are highly related to the interactive information among agents. The proposed framework can guide the development of more reliable and safer forecasting systems in the future.

Computational modeling of hemodynamics and risk of thrombosis in the left atrial appendage using patient-specific blood viscosity and boundary conditions at the mitral valve.

Hemodynamics play a vital role for the risk of thrombosis in the left atrial appendage (LAA) and left atrium (LA) for patients with atrial fibrillation. Accurate prediction of hemodynamics in the LA can provide important guidance for assessing the risk of thrombosis in the LAA. Patient specificity is a crucial factor in representing the true hemodynamic fields. In this study, we investigated the effects of blood rheology (as a function of hematocrit and shear rate), as well as patient-specific mitral valve (MV) boundary conditions (MV area and velocity profiles measured by ultrasound) on the hemodynamics and thrombosis potential of the LAA. Four scenarios were setup with different degrees of patient specificity. Though using a constant blood viscosity can classify the thrombus and non-thrombus patients for all the hemodynamic indicators, the risk of thrombosis was underestimated for all patients compared with patient-specific viscosities. The results with least patient specificities showed that patients prone to thrombosis predicted by three hemodynamic indicators were inconsistent with clinical observations. Moreover, though patients had the same MV inlet flow rate, different MV models lead to different trends in the risk of thrombosis in different patients. We also found that endothelial cell activation potential and relative residence time can effectively distinguish thrombus and non-thrombus patients for all the scenarios, relatively insensitive to patient specificities. Overall, the findings of this study provide useful insights on patients-specific hemodynamic simulations of the LA.

Depression and activities of daily living in elderly people applying for admission to public nursing facilities.

OBJECTIVE: The present study aims to (1) investigate the prevalence of positive screening for depression among elderly people applying for admission to public nursing facilities in Shenzhen and (2) to explore the correlation between depression and activities of daily living (ADL) among the elderly. METHODS: Visual and hearing acuity, ADL (obtained using the Barthel index), cognition and depression levels, and communication and social interaction skills were assessed in all elderly adults aged ≥60 years who applied for admission to public nursing facilities in Shenzhen between April 1, 2018, and December 31, 2019. RESULTS: A total of 1,113 elderly adults, 34.95% of whom were assessed as being depressed, were evaluated. Regarding dependence on the care of others, the ADL assessment results showed that 11.86% of the study subjects were fully dependent, 10.51% were severely dependent, 12.58% were moderately dependent, 42.86% were mildly dependent, and only 22.19% were fully capable of self-care. The univariate analysis suggested that gender, cognition level, visual and hearing acuity, and communication and social interaction skills were all correlated with the occurrence of depression. The prevalence of positive screening for depression was relatively high among subjects with an ADL score of ≤65; With an ADL score of ≤65, the prevalence of depression increased by 6% with every 1-point ADL score decrease. CONCLUSION: The prevalence of depression among elderly adults applying for admission to public nursing facilities in Shenzhen was high. For this reason, nursing facility caregivers should focus on screening elderly adults (especially subjects with impaired ADL function) for depression in order to improve their quality of life.

Influence of Inlet Boundary Conditions on the Prediction of Flow Field and Hemolysis in Blood Pumps Using Large-Eddy Simulation.

Inlet boundary conditions (BC) are one of the uncertainties which may influence the prediction of flow field and hemolysis in blood pumps. This study investigated the influence of inlet BC, including the length of inlet pipe, type of inlet BC (mass flow rate or experimental velocity profile) and turbulent intensity (no perturbation, 5%, 10%, 20%) on the prediction of flow field and hemolysis of a benchmark centrifugal blood pump (the FDA blood pump) and a commercial axial blood pump (Heartmate II), using large-eddy simulation. The results show that the influence of boundary conditions on integral pump performance metrics, including pressure head and hemolysis, is negligible. The influence on local flow structures, such as velocity distributions, mainly existed in the inlet. For the centrifugal FDA blood pump, the influence of type of inlet BC and inlet position on velocity distributions can also be observed at the diffuser. Overall, the effects of position of inlet and type of inlet BC need to be considered if local flow structures are the focus, while the influence of turbulent intensity is negligible and need not be accounted for during numerical simulations of blood pumps.

Simple and active magnetic-field stabilization for cold atom experiments.

Cold atom experiments usually need a controllable and low-noise bias magnetic field to provide a quantization axis. Most labs need home-made stabilization of the field according to the actual setup, as commercially available power supply cannot directly satisfy their requirements. Here, by measuring the field fluctuations and active feedback modulating current supply of the applied magnetic field, we successfully demonstrate a field of 10.58 G with a stability to the level of 2.8 × 10-7 in a duration of 5 min. The root mean square noise is reduced to 0.05 mG, compared to the noise of 1.3 mG without stabilization. The coherence time of the magnetic-field sensitive transition between the rubidium ground states F=1,mF=-1 and 1,0, as measured by Rabi oscillation, is extended to 19.2 ms from the unstabilized value of 1.3 ms. This result is long enough for most experiments on quantum simulation and precision measurement. As our system has no passive magnetic shielding and additional compensation coils, it is highly simple and compact to provide the stable magnetic field and would be adapted to various applications with cold atoms.

von Willebrand factor unfolding mediates platelet deposition in a model of high-shear thrombosis.

Thrombosis under high-shear conditions is mediated by the mechanosensitive blood glycoprotein von Willebrand factor (vWF). vWF unfolds in response to strong flow gradients and facilitates rapid recruitment of platelets in flowing blood. While the thrombogenic effect of vWF is well recognized, its conformational response in complex flows has largely been omitted from numerical models of thrombosis. We recently presented a continuum model for the unfolding of vWF, where we represented vWF transport and its flow-induced conformational change using convection-diffusion-reaction equations. Here, we incorporate the vWF component into our multi-constituent model of thrombosis, where the local concentration of stretched vWF amplifies the deposition rate of free-flowing platelets and reduces the shear cleaning of deposited platelets. We validate the model using three benchmarks: in vitro model of atherothrombosis, a stagnation point flow, and the PFA-100, a clinical blood test commonly used for screening for von Willebrand disease (vWD). The simulations reproduced the key aspects of vWF-mediated thrombosis observed in these experiments, such as the thrombus location, thrombus growth dynamics, and the effect of blocking platelet-vWF interactions. The PFA-100 simulations closely matched the reported occlusion times for normal blood and several hemostatic deficiencies, namely, thrombocytopenia, vWD type 1, and vWD type 3. Overall, this multi-constituent model of thrombosis enables macro-scale 3D simulations of thrombus formation in complex geometries over a wide range of shear rates and accounts for qualitative and quantitative hemostatic deficiencies in patient blood.

Rheological Properties of Organic Kerosene Gel Fuel.

Gel fuel potentially combines the advantages of solid fuel and liquid fuel due to its special rheological properties, which have essential impacts on the application of gel fuel in propulsion systems. In this paper, we study the rheological property of organic kerosene gel through a series of measurements on its viscosity as a function of the shear rate, temperature, and shear history. The measured datasets are then fitted with constitutive relationships between the viscosity and shear rate at three different levels: the power law shear-thinning model, the power law dependency on both the temperature and shear rate, and the thixotropic property. It is found that intense pre-shear could exhaust thixotropy and reduce viscosity of the kerosene gel. For the power law shear-thinning model, the consistency index increases with the gellant mass fraction, whereas the power law exponent remains constant. The dependence of viscosity on temperature could be well approximated by an empirical power law relationship. As for the thixotropic property of the kerosene gel, the fitted second-order kinetic model corresponds accurately to the viscosity at different shear rates and shear times. The constitutive models fitted in this work at different levels are consistent with each other and provide useful tools for further applications of organic kerosene gel fuel.

Study of Thixotropic Characteristics of a Kerosene Gel Propellant by Bayesian Optimization.

The rheological behavior of gel propellants is crucial for their practical applications, especially in the rocket engine and ramjet fields. The thixotropic characteristics of gel propellants are an important component of their rheological properties and have a notable impact on their flow and injection process. However, most gel propellants contain rich, dynamic cross-linked network structures, which impart complex non-Newtonian fluid properties, and it is difficult to establish a unified mathematical model. In view of this, this study addresses the thixotropy of a prepared RP-3 kerosene gel and determines the mathematical model and model parameters describing its thixotropy. Experiments show that the kerosene gel exhibits shear-thinning properties as well as thixotropy. To describe the microstructural changes in the gel, three thixotropic constitutive models are introduced to analyze the rheological data, and the constitutive equation parameters are optimized. The three models are all structural dynamic models, which can be used to describe microstructural changes within the material. In addition, the fitting of the constitutive equation is a multiparameter optimization problem, and an appropriate optimization method must be used for parameter fitting. Therefore, the Bayesian optimization method combined with Gaussian process regression and the upper confidence bound (UCB) acquisition function is used in the multiparameter fitting of the constitutive models. Both experiments and numerical results show that the thixotropic model, which introduces a pre-factor with shear strain and assumes that the breakdown of the gel structure is related to energy dissipation rather than the shear rate, has a better fitting effect and prediction ability with regard to the gel. Combined with transient experiments at different shear rates, the model parameters of the constitutive law can be determined quickly by applying the Bayesian optimization method.

Large eddy simulation as a fast and accurate engineering approach for the simulation of rotary blood pumps.

An accurate representation of the flow field in blood pumps is important for the design and optimization of blood pumps. The primary turbulence modeling methods applied to blood pumps have been the Reynolds-averaged Navier-Stokes (RANS) or URANS (unsteady RANS) method. Large eddy simulation (LES) method has been introduced to simulate blood pumps. Nonetheless, LES has not been widely used to assist in the design and optimization of blood pumps to date due to its formidable computational cost. The purpose of this study is to explore the potential of the LES technique as a fast and accurate engineering approach for the simulation of rotary blood pumps. The performance of "Light LES" (using the same time and spatial resolutions as the URANS) and LES in two rotary blood pumps was evaluated by comparing the results with the URANS and extensive experimental results. This study showed that the results of both "Light LES" and LES are superior to URANS, in terms of both performance curves and key flow features. URANS could not predict the flow separation and recirculation in diffusers for both pumps. In contrast, LES is superior to URANS in capturing these flows, performing well for both design and off-design conditions. The differences between the "Light LES" and LES results were relatively small. This study shows that with less computational cost than URANS, "Light LES" can be considered as a cost-effective engineering approach to assist in the design and optimization of rotary blood pumps.

A Continuum Model for the Unfolding of von Willebrand Factor.

von Willebrand Factor is a mechano-sensitive protein circulating in blood that mediates platelet adhesion to subendothelial collagen and platelet aggregation at high shear rates. Its hemostatic function and thrombogenic effect, as well as susceptibility to enzymatic cleavage, are regulated by a conformational change from a collapsed globular state to a stretched state. Therefore, it is essential to account for the conformation of the vWF multimers when modeling vWF-mediated thrombosis or vWF degradation. We introduce a continuum model of vWF unfolding that is developed within the framework of our multi-constituent model of platelet-mediated thrombosis. The model considers two interconvertible vWF species corresponding to the collapsed and stretched conformational states. vWF unfolding takes place via two regimes: tumbling in simple shear and strong unfolding in flows with dominant extensional component. These two regimes were demonstrated in a Couette flow between parallel plates and an extensional flow in a cross-slot geometry. The vWF unfolding model was then verified in several microfluidic systems designed for inducing high-shear vWF-mediated thrombosis and screening for von Willebrand Disease. The model predicted high concentration of stretched vWF in key regions where occlusive thrombosis was observed experimentally. Strong unfolding caused by the extensional flow was limited to the center axis or middle plane of the channels, whereas vWF unfolding near the channel walls relied upon the shear tumbling mechanism. The continuum model of vWF unfolding presented in this work can be employed in numerical simulations of vWF-mediated thrombosis or vWF degradation in complex geometries. However, extending the model to 3-D arbitrary flows and turbulent flows will pose considerable challenges.

On the Optimization of a Centrifugal Maglev Blood Pump Through Design Variations.

Centrifugal blood pumps are usually designed with secondary flow paths to avoid flow dead zones and reduce the risk of thrombosis. Due to the secondary flow path, the intensity of secondary flows and turbulence in centrifugal blood pumps is generally very high. Conventional design theory is no longer applicable to centrifugal blood pumps with a secondary flow path. Empirical relationships between design variables and performance metrics generally do not exist for this type of blood pump. To date, little scientific study has been published concerning optimization and experimental validation of centrifugal blood pumps with secondary flow paths. Moreover, current hemolysis models are inadequate in an accurate prediction of hemolysis in turbulence. The purpose of this study is to optimize the hydraulic and hemolytic performance of an inhouse centrifugal maglev blood pump with a secondary flow path through variation of major design variables, with a focus on bringing down intensity of turbulence and secondary flows. Starting from a baseline design, through changing design variables such as blade angles, blade thickness, and position of splitter blades. Turbulent intensities have been greatly reduced, the hydraulic and hemolytic performance of the pump model was considerably improved. Computational fluid dynamics (CFD) combined with hemolysis models were mainly used for the evaluation of pump performance. A hydraulic test was conducted to validate the CFD regarding the hydraulic performance. Collectively, these results shed light on the impact of major design variables on the performance of modern centrifugal blood pumps with a secondary flow path.

Influence of shear rate and surface chemistry on thrombus formation in micro-crevice.

Thromboembolic complications remain a central issue in management of patients on mechanical circulatory support. Despite the best practices employed in design and manufacturing of modern ventricular assist devices, complexity and modular nature of these systems often introduces internal steps and crevices in the flow path which can serve as nidus for thrombus formation. Thrombotic potential is influenced by multiple factors including the characteristics of the flow and surface chemistry of the biomaterial. This study explored these elements in the setting of blood flow over a micro-crevice using a multi-constituent numerical model of thrombosis. The simulations reproduced the platelet deposition patterns observed experimentally and elucidated the role of flow, shear rate, and surface chemistry in shaping the deposition. The results offer insights for design and operation of blood-contacting devices.

Simulation of thrombosis in a stenotic microchannel: The effects of vWF-enhanced shear activation of platelets.

This study was undertaken to develop a numerical/computational simulation of von Willebrand Factor (vWF) - mediated platelet shear activation and deposition in an idealized stenosis. Blood is treated as a multi-constituent mixture comprised of a linear fluid component and a porous solid component (thrombus). Chemical and biological species involved in coagulation are modeled using a system of coupled convection-reaction-diffusion (CRD) equations. This study considers the cumulative effect of shear stress (history) on platelet activation. The vWF activity is modeled as an enhancement function for the shear stress accumulation and is related to the experimentally-observed unfolding rate of vWF. A series of simulations were performed in an idealized stenosis in which the predicted platelets deposition agreed well with previous experimental observations spatially and temporally, including the reduction of platelet deposition with decreasing expansion angle. Further simulation indicated a direct relationship between vWF-mediated platelet deposition and degree of stenosis. Based on the success with these benchmark simulations, it is hoped that the model presented here may provide additional insight into vWF-mediated thrombosis and prove useful for the development of more hemo-compatible blood-wetted devices in the future.

Simulation of blood flow in a sudden expansion channel and a coronary artery.

In this paper, we numerically simulate the flow of blood in two benchmark problems: the flow in a sudden expansion channel and the flow through an idealized curved coronary artery with pulsatile inlet velocity. Blood is modeled as a suspension (a non-linear complex fluid) and the movement of the red blood cell (RBCs) is modeled by using a concentration flux equation. The viscosity of blood is obtained from experimental data. In the sudden expansion flow, the predicted velocity profiles for two different Reynolds numbers (based on the inlet velocity) agree well with the available experiments; furthermore, the numerical results also show that after the sudden expansion there exists a RBCs depletion region. For the second problem, the idealized curved coronary artery, it is found that the RBCs move towards and concentrate near the inner surface where the viscosity is higher and the shear stress lower; this phenomenon may be related to the atherosclerotic plaque formation which usually occurs on the inside surface of the arteries.

A non-linear fluid suspension model for blood flow.

Motivated by the complex rheological behaviors observed in small/micro scale blood vessels, such as the Fahraeus effect, plasma-skimming, shear-thinning, etc., we develop a non-linear suspension model for blood. The viscosity is assumed to depend on the volume fraction (hematocrit) and the shear rate. The migration of the red blood cells (RBCs) is studied using a concentration flux equation. A parametric study with two representative problems, namely simple shear flow and a pressure driven flow demonstrate the ability of this reduced-order model to reproduce several key features of the two-fluid model (mixture theory approach), with much lower computational cost.

Development of a High-Throughput Magnetic Separation Device for Malaria-Infected Erythrocytes.

This study describes a non-dilutive high-gradient magnetic separation (HGMS) device intended to continuously remove malaria-infected red blood cells (iRBCs) from the circulation. A mesoscale prototype device with disposable photo-etched ferromagnetic grid and reusable permanent magnet was designed with a computationally-optimized magnetic force. The prototype device was evaluated in vitro using a non-pathogenic analog for malaria-infected blood, comprised of 24% healthy RBCs, 6% human methemoglobin RBCs (metRBCs), and 70% phosphate buffer solution (PBS). The device provided a 27.0 ± 2.2% reduction of metRBCs in a single pass at a flow rate of 77 µL min-1. This represents a clearance rate over 380 times greater throughput than microfluidic devices reported previously. These positive results encourage development of a clinical scale system that would economize time and donor blood for treating severe malaria.

Multi-Constituent Simulation of Thrombus Deposition.

In this paper, we present a spatio-temporal mathematical model for simulating the formation and growth of a thrombus. Blood is treated as a multi-constituent mixture comprised of a linear fluid phase and a thrombus (solid) phase. The transport and reactions of 10 chemical and biological species are incorporated using a system of coupled convection-reaction-diffusion (CRD) equations to represent three processes in thrombus formation: initiation, propagation and stabilization. Computational fluid dynamic (CFD) simulations using the libraries of OpenFOAM were performed for two illustrative benchmark problems: in vivo thrombus growth in an injured blood vessel and in vitro thrombus deposition in micro-channels (1.5 mm × 1.6 mm × 0.1 mm) with small crevices (125 µm × 75 µm and 125 µm × 137 µm). For both problems, the simulated thrombus deposition agreed very well with experimental observations, both spatially and temporally. Based on the success with these two benchmark problems, which have very different flow conditions and biological environments, we believe that the current model will provide useful insight into the genesis of thrombosis in blood-wetted devices, and provide a tool for the design of less thrombogenic devices.

High fidelity computational simulation of thrombus formation in Thoratec HeartMate II continuous flow ventricular assist device.

Continuous flow ventricular assist devices (cfVADs) provide a life-saving therapy for severe heart failure. However, in recent years, the incidence of device-related thrombosis (resulting in stroke, device-exchange surgery or premature death) has been increasing dramatically, which has alarmed both the medical community and the FDA. The objective of this study was to gain improved understanding of the initiation and progression of thrombosis in one of the most commonly used cfVADs, the Thoratec HeartMate II. A computational fluid dynamics simulation (CFD) was performed using our recently updated mathematical model of thrombosis. The patterns of deposition predicted by simulation agreed well with clinical observations. Furthermore, thrombus accumulation was found to increase with decreased flow rate, and can be completely suppressed by the application of anticoagulants and/or improvement of surface chemistry. To our knowledge, this is the first simulation to explicitly model the processes of platelet deposition and thrombus growth in a continuous flow blood pump and thereby replicate patterns of deposition observed clinically. The use of this simulation tool over a range of hemodynamic, hematological, and anticoagulation conditions could assist physicians to personalize clinical management to mitigate the risk of thrombosis. It may also contribute to the design of future VADs that are less thrombogenic.

Design of microfluidic channels for magnetic separation of malaria-infected red blood cells.

This study is motivated by the development of a blood cell filtration device for removal of malaria-infected, parasitized red blood cells (pRBCs). The blood was modeled as a multi-component fluid using the computational fluid dynamics discrete element method (CFD-DEM), wherein plasma was treated as a Newtonian fluid and the red blood cells (RBCs) were modeled as soft-sphere solid particles which move under the influence of drag, collisions with other RBCs, and a magnetic force. The CFD-DEM model was first validated by a comparison with experimental data from Han et al. 2006 (Han and Frazier 2006) involving a microfluidic magnetophoretic separator for paramagnetic deoxygenated blood cells. The computational model was then applied to a parametric study of a parallel-plate separator having hematocrit of 40% with a 10% of the RBCs as pRBCs. Specifically, we investigated the hypothesis of introducing an upstream constriction to the channel to divert the magnetic cells within the near-wall layer where the magnetic force is greatest. Simulations compared the efficacy of various geometries upon the stratification efficiency of the pRBCs. For a channel with nominal height of 100 µm, the addition of an upstream constriction of 80% improved the proportion of pRBCs retained adjacent to the magnetic wall (separation efficiency) by almost 2 fold, from 26% to 49%. Further addition of a downstream diffuser reduced remixing, hence improved separation efficiency to 72%. The constriction introduced a greater pressure drop (from 17 to 495 Pa), which should be considered when scaling-up this design for a clinical-sized system. Overall, the advantages of this design include its ability to accommodate physiological hematocrit and high throughput - which is critical for clinical implementation as a blood-filtration system.

Visualization and analysis of biomaterial-centered thrombus formation within a defined crevice under flow.

The blood flow pathway within a device, together with the biomaterial surfaces and status of the patient's blood, are well-recognized factors in the development of thrombotic deposition and subsequent embolization. Blood flow patterns are of particular concern for devices such as blood pumps (i.e. ventricular assist devices, VADs) where shearing forces can be high, volumes are relatively large, and the flow fields can be complex. However, few studies have examined the effect of geometric irregularities on thrombus formation on clinically relevant opaque materials under flow. The objective of this study was to quantify human platelet deposition onto Ti6Al4V alloys, as well as positive and negative control surfaces, in the region of defined crevices (∼50-150 µm in width) that might be encountered in many VADs or other cardiovascular devices. To achieve this, reconstituted fresh human blood with hemoglobin-depleted red blood cells (to achieve optical clarity while maintaining relevant rheology), long working optics, and a custom designed parallel plate flow chamber were employed. The results showed that the least amount of platelet deposition occurred in the largest crevice size examined, which was counterintuitive. The greatest levels of deposition occurred in the 90 µm and 53 µm crevices at the lower wall shear rate. The results suggest that while crevices may be unavoidable in device manufacturing, the crevice size might be tailored, depending on the flow conditions, to reduce the risk of thromboembolic events. Further, these data might be used to improve the accuracy of predictive models of thrombotic deposition in cardiovascular devices to help optimize the blood flow path and reduce device thrombogenicity.