Image-based predictive modelling frameworks for personalised drug delivery in cancer therapy.

Personalised drug delivery enables a tailored treatment plan for each patient compared to conventional drug delivery, where a generic strategy is commonly employed. It can not only achieve precise treatment to improve effectiveness but also reduce the risk of adverse effects to improve patients' quality of life. Drug delivery involves multiple interconnected physiological and physicochemical processes, which span a wide range of time and length scales. How to consider the impact of individual differences on these processes becomes critical. Multiphysics models are an open system that allows well-controlled studies on the individual and combined effects of influencing factors on drug delivery outcomes while accommodating the patient-specific in vivo environment, which is not economically feasible through experimental means. Extensive modelling frameworks have been developed to reveal the underlying mechanisms of drug delivery and optimise effective delivery plans. This review provides an overview of currently available models, their integration with advanced medical imaging modalities, and code packages for personalised drug delivery. The potential to incorporate new technologies (i.e., machine learning) in this field is also addressed for development.

In silico study of combination thrombolytic therapy with alteplase and mutant pro-urokinase for fibrinolysis in ischemic stroke.

The synergistic advantage of combining tissue plasminogen activator (tPA) with pro-urokinase (proUK) for thrombolysis has been demonstrated in several in vitro experiments, and a single site proUK mutant (m-proUK) has been developed for better stability in plasma. Based on these studies, combination thrombolytic therapy with intravenous tPA and m-proUK has been suggested as a promising treatment for patients with ischemic stroke. This paper evaluates the efficacy and safety of the dual therapy by computational simulations of pharmacokinetics and pharmacodynamics coupled with a local fibrinolysis model. Seven dose regimens are simulated and compared with the standard intravenous tPA monotherapy. Our simulation results provide more insights into the complementary reaction mechanisms of tPA and m-proUK during clot lysis and demonstrate that the dual therapy can achieve a similar recanalization time (about 50 min) to tPA monotherapy, while keeping the circulating fibrinogen level within a normal range. Specifically, our results show that for all dual therapies with a 5 mg tPA bolus, the plasma concentration of fibrinogen remains stable at around 7.5 µM after a slow depletion over 50 min, whereas a rapid depletion of circulating fibrinogen (to 5 µM) is observed with the standard tPA therapy, indicating the potential advantage of dual therapy in reducing the risk of intracranial hemorrhage. Through simulations of varying dose combinations, it has been found that increasing tPA bolus can significantly affect fibrinogen level but only moderately improves recanalization time. Conversely, m-proUK doses and infusion duration exhibit a mild impact on fibrinogen level but significantly affect recanalization time. Therefore, future optimization of dose regimen should focus on limiting the tPA bolus while adjusting m-proUK dosage and infusion rate. Such adjustments could potentially maximize the therapeutic advantages of this combination therapy for ischemic stroke treatment.

In Silico Study of Different Thrombolytic Agents for Fibrinolysis in Acute Ischemic Stroke.

Alteplase is the only FDA-approved drug for thrombolysis in acute ischemic stroke (AIS). Meanwhile, several thrombolytic drugs are deemed to be promising candidates to substitute alteplase. This paper evaluates the efficacy and safety of urokinase, ateplase, tenecteplase, and reteplase for intravenous AIS therapy by computational simulations of the pharmacokinetics and pharmacodynamics combined with a local fibrinolysis model. The performances of the drugs are evaluated by comparing clot lysis time, plasminogen activator inhibitor (PAI) inhibition resistance, intracranial hemorrhage (ICH) risk, and activation time from drug administration to clot lysis. Our results reveal that urokinase has the quickest lysis completion but the highest ICH risk due to excess fibrinogen depletion in systemic plasma. While tenecteplase and alteplase have very similar thrombolysis efficacy, tenecteplase has a lower risk of ICH and better resistance to PAI-1. Among the four simulated drugs, reteplase has the slowest fibrinolysis rate, but fibrinogen concentration in systemic plasma is unaffected during thrombolysis.

Engineered multi-scale roughness of carbon nanofiller-embedded 3D printed spacers for membrane distillation.

Membrane distillation (MD) transfers heat and mass simultaneously through a hydrophobic membrane. Hence, it is sensitive to both concentration and temperature polarisation (CP and TP) effects. In this study, we fabricated feed spacers to improve MD efficiency by alleviating the polarisation effects. First, a 3D printed spacer design was optimised to show superior performance amongst the others tested. Then, to further enhance spacer performance, we incorporated highly thermally stable carbon nanofillers, including carbon nanotubes (CNT) and graphene, in the fabrication of filaments for 3D printing. All the fabricated spacers had a degree of engineered multi-scale roughness, which was relatively high compared to that of the polylactic acid (PLA) spacer (control). The use of nanomaterial-incorporated spacers increased the mean permeate flux significantly compared to the PLA spacer (27.1 L/m2h (LMH)): a 43% and 75% increase when using the 1% graphene-incorporated spacer (38.9 LMH) and 2% CNT incorporated spacer (47.5 LMH), respectively. This could be attributed to the locally enhanced turbulence owing to the multi-scale roughness formed on the spacer, which further increased the vaporisation rate through the membrane. Interestingly, only the CNT-embedded spacer markedly reduced the ion permeation through the membrane, which may be due to the effective reduction of CP. This further decreased with increasing CNT concentration, confirming that the CNT spacer can simultaneously reduce the CP and TP effects in the MD process. Finally, we successfully proved that the multi-scale roughness of the spacer surface induces micromixing near the membrane walls, which can improve the MD performance via computational fluid dynamics.

Simulation Model for Prediction of Gas Separation in Membrane Contactor Process.

The purpose of this study is to establish a practical simulation model based on mass balance, mass transport equations and equilibrium equation between gas and liquid phases across a porous membrane in membrane contactor process in order to predict the separation behavior by the gassing process of gas mixture in membrane contactor. The established simulation model was verified by comparison between the simulated values and real process values in the separation of CH4/CO2 mixture, showing an excellent agreement between them. The parameter R-value in the model, which is a kind of the permeability of permeant across porous membrane, has been determined by fitting a numerical solution of the model equation to the experimental data to obtain a practical value of the parameter. A parametric study on the gassing process of N2/CO2 mixture in membrane contactor was made with the help of the practical simulation model to investigate the effects of operation parameters on separation performance and to characterize the separation behavior of membrane contactor process. A series of simulations of the separation of N2/CO2 mixture in membrane contactor were conducted, and the optimization on the membrane process was discussed to maximize the separation performance in terms of N2 recovery percent in retentate and CO2 permeation rate. It was observed from the analysis of the result of the simulation that liquid flow rate has a negative effect on N2 recovery percent in retentate but a positive effect on the separation of CO2, while R-value affects the separation performance in the other way. It is confirmed in this study that the developed simulation can be used as a tool to optimize the parameters, i.e., feed gas pressure, liquid flow rate and R-value to maximize the separation performance.

Multiphysics Modelling and Simulation of Thrombolysis via Activated Platelet-Targeted Nanomedicine.

PURPOSE: This study establishes a multiphysics simulation platform for both conventional and targeted thrombolysis using tissue plasminogen activator (tPA). Based on our computational results, the effects of therapeutic parameters on the dynamics of thrombolysis and the risk of side effects are investigated. METHODS: The model extends our previously developed one-dimensional(1D) mathematical models for fibrinolysis by incorporating targeted thrombolysis. It consists of two parts: (i) a coupled mathematical model of systemic pharmacokinetics (PK) and pharmacodynamics (PD) and local PD in a 1D occluded artery, and (ii) a mechanistic model for a targeted thrombolytic system via activated platelet-targeted tPA-loaded nanovesicles (tPA-NV), with model parameters derived from our in vitro experiments. A total of 16 therapeutic scenarios are simulated by varying the clot location and composition as well as the dosing regimen with free tPA or tPA-NV. RESULTS: Our simulation results indicate that tPA-NV offers several advantages over free tPA for thrombolysis. It reduces systemic exposure of tPA, thereby minimising the risk of bleeding complications. Simulations with different tPA-NV doses reveal that tPA-NV at 10% of the recommended dose can be as effective as the standard regimen with the full recommended dose of free tPA, demonstrating the potential of our tPA-NV as a new thrombolytic strategy with a reduced tPA dose. Moreover, faster recanalisation can be achieved with tPA-NV, especially for platelet-rich(or fibrin-poor) clots. CONCLUSIONS: Our simulation platform for thrombolysis with well-tuned model parameters can be used to evaluate and optimise treatment regimens of existing and new thrombolytic therapies via benefit/risk assessment under various therapeutic scenarios.

An integrated fluid-structure interaction and thrombosis model for type B aortic dissection.

False lumen thrombosis (FLT) in type B aortic dissection has been associated with the progression of dissection and treatment outcome. Existing computational models mostly assume rigid wall behavior which ignores the effect of flap motion on flow and thrombus formation within the FL. In this study, we have combined a fully coupled fluid-structure interaction (FSI) approach with a shear-driven thrombosis model described by a series of convection-diffusion reaction equations. The integrated FSI-thrombosis model has been applied to an idealized dissection geometry to investigate the interaction between vessel wall motion and growing thrombus. Our simulation results show that wall compliance and flap motion can influence the progression of FLT. The main difference between the rigid and FSI models is the continuous development of vortices near the tears caused by drastic flap motion up to 4.45 mm. Flap-induced high shear stress and shear rates around tears help to transport activated platelets further to the neighboring region, thus speeding up thrombus formation during the accelerated phase in the FSI models. Reducing flap mobility by increasing the Young's modulus of the flap slows down the thrombus growth. Compared to the rigid model, the predicted thrombus volume is 25% larger using the FSI-thrombosis model with a relatively mobile flap. Furthermore, our FSI-thrombosis model can capture the gradual effect of thrombus growth on the flow field, leading to flow obstruction in the FL, increased blood viscosity and reduced flap motion. This model is a step closer toward simulating realistic thrombus growth in aortic dissection, by taking into account the effect of intimal flap and vessel wall motion.

Modelling Combined Intravenous Thrombolysis and Mechanical Thrombectomy in Acute Ischaemic Stroke: Understanding the Relationship between Stent Retriever Configuration and Clot Lysis Mechanisms.

Background: Combined intravenous thrombolysis and mechanical thrombectomy (IVT-MT) is a common treatment in acute ischaemic stroke, however the interaction between IVT and MT from a physiological standpoint is poorly understood. In this pilot study, we conduct numerical simulations of combined IVT-MT with various idealised stent retriever configurations to evaluate performance in terms of complete recanalisation times and lysis patterns. Methods: A 3D patient-specific geometry of a terminal internal carotid artery with anterior and middle cerebral arteries is reconstructed, and a thrombus is artificially implanted in the MCA branch. Various idealised stent retriever configurations are implemented by varying stent diameter and stent placement, and a configuration without a stent retriever provides a baseline for comparison. A previously validated multi-level model of thrombolysis is used, which incorporates blood flow, drug transport, and fibrinolytic reactions within a fibrin thrombus. Results: Fastest total recanalisation was achieved in the thrombus without a stent retriever, with lysis times increasing with stent retriever diameter. Two mechanisms of clot lysis were established: axial and radial permeation. Axial permeation from the clot front was the primary mechanism of lysis in all configurations, as it facilitated increased protein binding with fibrin fibres. Introducing a stent retriever channel allowed for radial permeation, which occurred at the fluid-thrombus interface, although lysis was much slower in the radial direction because of weaker secondary velocities. Conclusions: Numerical models can be used to better understand the complex physiological relationship between IVT and MT. Two different mechanisms of lysis were established, providing a basis towards improving the efficacy of combined treatments.

Fibrinogen-mimicking, multiarm nanovesicles for human thrombus-specific delivery of tissue plasminogen activator and targeted thrombolytic therapy.

Clinical use of tissue plasminogen activator (tPA) in thrombolytic therapy is limited by its short circulation time and hemorrhagic side effects. Inspired by fibrinogen binding to activated platelets, we report a fibrinogen-mimicking, multiarm nanovesicle for thrombus-specific tPA delivery and targeted thrombolysis. This biomimetic system is based on the lipid nanovesicle coated with polyethylene glycol (PEG) terminally conjugated with a cyclic RGD (cRGD) peptide. Our experiments with human blood demonstrated its highly selective binding to activated platelets and efficient tPA release at a thrombus site under both static and physiological flow conditions. Its clot dissolution time in a microfluidic system was comparable to that of free tPA. Furthermore, we report a purpose-built computational model capable of simulating targeted thrombolysis of the tPA-loaded nanovesicle and with a potential in predicting the dynamics of thrombolysis in physiologically realistic scenarios. This combined experimental and computational work presents a promising platform for development of thrombolytic nanomedicines.

Correlations for Concentration Polarization and Pressure Drop in Spacer-Filled RO Membrane Modules Based on CFD Simulations.

Empirical correlations for mass transfer coefficient and friction factor are often used in process models for reverse osmosis (RO) membrane systems. These usually involve four dimensionless groups, namely Reynolds number (Re), Sherwood number (Sh), friction factor (f), and Schmidt number (Sc), with the associated coefficients and exponents being obtained by fitting to experimental data. However, the range of geometric and operating conditions covered by the experiments is often limited. In this study, new dimensionless correlations for concentration polarization (CP) modulus and friction factor are presented, which are obtained by dimensional analysis and using simulation data from computational fluid dynamics (CFD). Two-dimensional CFD simulations are performed on three configurations of spacer-filled channels with 76 combinations of operating and geometric conditions for each configuration, covering a broad range of conditions encountered in RO membrane systems. Results obtained with the new correlations are compared with those from existing correlations in the literature. There is good consistency in the predicted CP with mean discrepancies less than 6%, but larger discrepancies for pressure gradient are found among the various friction factor correlations. Furthermore, the new correlations are implemented in a process model with six spiral wound modules in series and the predicted recovery, pressure drop, and specific energy consumption are compared with a reference case obtained by ROSA (Reverse Osmosis System Analysis, The Dow Chemical Company). Differences in predicted recovery and pressure drop are up to 5% and 83%, respectively, highlighting the need for careful selection of correlations when using predictive models in process design. Compared to existing mass transfer correlations, a distinct advantage of our correlations for CP modulus is that they can be directly used to estimate the impact of permeate flux on CP at a membrane surface without having to resort to the film theory.

Effect of intimal flap motion on flow in acute type B aortic dissection by using fluid-structure interaction.

A monolithic, fully coupled fluid-structure interaction (FSI) computational framework was developed to account for dissection flap motion in acute type B aortic dissection (TBAD). Analysis of results included wall deformation, pressure, flow, wall shear stress (WSS), von Mises stress and comparison of hemodynamics between rigid wall and FSI models. Our FSI model mimicked realistic wall deformation that resulted in maximum compression of the distal true lumen (TL) by 21.4%. The substantial movement of intimal flap mostly affected flow conditions in the false lumen (FL). Flap motion facilitated more flow entering the FL at peak systole, with the TL to FL flow split changing from 88:12 in the rigid model to 83:17 in the FSI model. There was more disturbed flow in the FL during systole (5.8% FSI vs 5.2% rigid) and diastole (13.5% FSI vs 9.8% rigid), via a λ2 -criterion. The flap-induced disturbed flow near the tears in the FSI model caused an increase of local WSS by up to 70.0% during diastole. This resulted in a significant reduction in the size of low time-averaged WSS (TAWSS) regions in the FL (113.11 cm2 FSI vs 177.44 cm2 rigid). Moreover, the FSI model predicted lower systolic pressure, higher diastolic pressure, and hence lower pulse pressure. Our results provided new insights into the possible impact of flap motion on flow in aortic dissections, which are particularly important when evaluating hemodynamics of acute TBAD. NOVELTY STATEMENT: Our monolithic fully coupled FSI computational framework is able to reproduce experimentally measured range of flap deformation in aortic dissection, thereby providing novel insights into the influence of physiological flap motion on the flow and pressure distributions. The drastic flap movement increases the flow resistance in the true lumen and causes more disturbed flow in the false lumen, as visualized through the λ2 criterion. The flap-induced luminal pressure is dampened, thereby affecting pressure measures, which may serve as potential prognostic indicators for late complications in acute uncomplicated TBAD patients.

Thermosensitive Liposome-Mediated Drug Delivery in Chemotherapy: Mathematical Modelling for Spatio-temporal Drug Distribution and Model-Based Optimisation.

Thermosensitive liposome-mediated drug delivery has shown promising results in terms of improved therapeutic efficacy and reduced side effects compared to conventional chemotherapeutics. In order to facilitate our understanding of the transport mechanisms and their complex interplays in the drug delivery process, computational models have been developed to simulate the multiple steps involved in liposomal drug delivery to solid tumours. In this study we employ a multicompartmental model for drug-loaded thermosensitive liposomes, with an aim to identify the key transport parameters in determining therapeutic dosing and outcomes. The computational model allows us to not only examine the temporal and spatial variations of drug concentrations in the different compartments by utilising the tumour cord concept, but also assess the therapeutic efficacy and toxicity. In addition, the influences of key factors on systemic plasma concentration and intracellular concentration of the active drug are investigated; these include different chemotherapy drugs, release rate constants and heating duration. Our results show complex relationships between these factors and the predicted therapeutic outcome, making it difficult to identify the "best" parameter set. To overcome this challenge, a model-based optimisation method is proposed in an attempt to find a set of release rate constants and heating duration that can maximise intracellular drug concentration while minimising systemic drug concentration. Optimisation results reveal that under the operating conditions and ranges examined, the best outcome would be achieved with a low drug release rate at physiological temperature, combined with a moderate to high release rate at mild hyperthermia and 1 h heating after injection.

Computational simulations of thrombolysis in acute stroke: Effect of clot size and location on recanalisation.

Acute ischaemic stroke can be treated by intravenous thrombolysis whereby tissue plasminogen activator (tPA) is infused to dissolve clots that block blood supply to the brain. In this study, we aim to examine the influence of clot location and size on lysis pattern and recanalisation by using a recently developed computational modelling framework for thrombolysis under physiological flow conditions. An image-based patient-specific model is reconstructed which consists of the internal carotid bifurcation with the A1 segment of anterior cerebral arteries and M1 segment of middle cerebral arteries, and the M1 bifurcation containing the M2 segments. By varying the clot size and location, 7 scenarios are simulated mimicking thrombolysis of M1 and M2 occlusions. Our results show that initial breakthrough always occurs along the inner curvature of the occluded cerebral artery, due to prolonged tPA residence time in the recirculation zone. For a given occlusion site, lysis completion time appears to increase almost quadratically with the initial clot volume; whereas for a given clot volume, the simulated M2 occlusions take up to 30% longer for complete lysis compared to the corresponding M1 occlusions.

Mathematical Modelling of Intravenous Thrombolysis in Acute Ischaemic stroke: Effects of Dose Regimens on Levels of Fibrinolytic Proteins and Clot Lysis Time.

Thrombolytic therapy is one of the medical procedures in the treatment of acute ischaemic stroke (AIS), whereby the tissue plasminogen activator (tPA) is intravenously administered to dissolve the obstructive blood clot. The treatment of AIS by thrombolysis can sometimes be ineffective and it can cause serious complications, such as intracranial haemorrhage (ICH). In this study, we propose an efficient mathematical modelling approach that can be used to evaluate the therapeutic efficacy and safety of thrombolysis in various clinically relevant scenarios. Our model combines the pharmacokinetics and pharmacodynamics of tPA with local clot lysis dynamics. By varying the drug dose, bolus-infusion delay time, and bolus-infusion ratio, with the FDA approved dosing protocol serving as a reference, we have used the model to simulate 13 dose regimens. Simulation results are compared for temporal concentrations of fibrinolytic proteins in plasma and the time that is taken to achieve recanalisation. Our results show that high infusion rates can cause the rapid degradation of plasma fibrinogen, indicative of increased risk for ICH, but they do not necessarily lead to fast recanalisation. In addition, a bolus-infusion delay results in an immediate drop in plasma tPA concentration, which prolongs the time to achieve recanalisation. Therefore, an optimal administration regimen should be sought by keeping the tPA level sufficiently high throughout the treatment and maximising the lysis rate while also limiting the degradation of fibrinogen in systemic plasma. This can be achieved through model-based optimisation in the future.

An activated-platelet-sensitive nanocarrier enables targeted delivery of tissue plasminogen activator for effective thrombolytic therapy.

It remains a major challenge to develop a selective and effective fibrinolytic system for thrombolysis with minimal undesirable side effects. Herein, we report a multifunctional liposomal system (164.6 ±â€¯5.3 nm in diameter) which can address this challenge through targeted delivery and controlled release of tissue plasminogen activator (tPA) at the thrombus site. The tPA-loaded liposomes were PEGylated to improve their stability, and surface coated with a conformationally-constrained, cyclic arginine-glycine-aspartic acid (cRGD) to enable highly selective binding to activated platelets. The in vitro drug release profiles at 37 °C showed that over 90% of tPA was released through liposomal membrane destabilization involving membrane fusion upon incubation with activated platelets within 1 h, whereas passive release of the encapsulated tPA in pH 7.4 PBS buffer was 10% after 6 h. The release of tPA could be readily manipulated by changing the concentration of activated platelets. The presence of activated platelets enabled the tPA-loaded, cRGD-coated, PEGylated liposomes to induce efficient fibrin clot lysis in a fibrin-agar plate model and the encapsulated tPA retained 97.4 ±â€¯1.7% of fibrinolytic activity as compared with that of native tPA. Furthermore, almost complete blood clot lysis was achieved in 75 min, showing considerably higher and quicker thrombolytic activity compared to the tPA-loaded liposomes without cRGD labelling. These results suggest that the nano-sized, activated-platelet-sensitive, multifunctional liposomes could facilitate selective delivery and effective release of tPA at the site of thrombus, thus achieving efficient clot dissolution whilst minimising undesirable side effects.

Computational Simulations of Thrombolytic Therapy in Acute Ischaemic Stroke.

Ischaemic stroke can occur when an artery to the brain is blocked by a blood clot. The use of thrombolytic agents, such as tissue plasminogen activator (tPA), to dissolve the occluding clot is limited by the risk of intracerebral haemorrhage (ICH), a known side effect associated with tPA. We developed a computational thrombolysis model for a 3D patient-specific artery coupled with a compartmental model for temporal concentrations of tPA and lysis proteins during intravenous infusion of tPA, in order to evaluate the effects of tPA dose on the efficacy of thrombolytic therapy and the risk of ICH. The model was applied to a 3-mm-long fibrin clot with two different fibrin fibre radii in the middle cerebral artery (MCA) - a setting relevant to ischaemic stroke, and results for different tPA dose levels and fibrin fibre radii were compared. Our simulation results showed that clot lysis was accelerated at higher tPA doses at the expense of a substantial increase in the risk of ICH. It was also found that a fine clot with a smaller fibre radius dissolved much slowly than a coarse clot due to a slower tPA penetration into the clots.