OpenUCT :: Browsing by Author "Ngoepe, Malebogo"

Browsing by Author "Ngoepe, Malebogo"

Now showing 1 - 13 of 13

Open Access
A Mechano-Chemical Computational model of Deep Vein Thrombosis
(2021) Jimoh-Taiwo, Qudus Boluwatife; Ngoepe, Malebogo
Deep Vein Thrombosis (DVT) is the formation of a blood clot in a vein, usually in the body's lower extremities. If untreated, DVT can lead to pulmonary embolism (PE), heart attack and/or stroke, which can be fatal. According to literature, DVT affects 0.2% of people in developed countries and about 0.3%-1% in developing countries. In the past, various computational models of DVT were developed. Most models account for either the mechanical factors or biochemical factors involved with DVT. Developing a model that accounts for both factors will improve our understanding of the coagulation process. This study developed a three-dimensional DVT computational model in idealized and realistic common femoral vein (CFV) geometries. The model considers the biochemical reactions between thrombin and fibrinogen, pulsatile blood flow, and clot growth within the vessel. The model was validated using a simplified experimental setup with flow, thrombin, and fibrinogen. Computational fluid dynamics (CFD) simulations were carried out using the ANSYS modelling suite. The Navier-Stokes equations were solved to determine the fluid flow. Based on a clinical dataset of pulsatile blood flow, the laminar flow of blood with a Poiseuille velocity profile was applied at the inlet. Darcy's law was used to account for porosity changes in the clot, with the clot represented by zones with lower porosities. The transport equations were used for changes in the concentration of the biochemical protein species. Thrombin was released into the bloodstream from an injury zone on the wall of the vein. The Michaelis-Menten equation was used to represent the conversion of thrombin and fibrinogen to fibrin, the final product of the coagulation process. The computational model solves the blood flow pattern proximally, locally, and distally to clot formation at the injury zone. The model also predicts the size of the clot and the rate of clot growth. The model was first developed in a two-dimensional geometry. This model was used to investigate clot formation under different cases comparing how introducing thrombin as a flux value differs from specifying it as a fixed concentration. It was confirmed that to apply the flux condition, the thrombin concentration needs to be divided by a factor derived by multiplying the area of the injury zone and the time step size. The same model was then used to conduct a parametric study to determine the effects of varying parameters such as inlet velocity, vein diameter, and peak thrombin concentration on the size and shape of clot formed. Peak thrombin concentration was the key factor driving the initiation and propagation of clot in the vein. The model was then extended to an idealized three-dimensional geometry. This computational model was validated using results from an experimental clot growth study. The experiment comprised a steady flow of fibrinogen in a cylindrical pipe, with an injection of thrombin into the flow at the injury site, resulting in fibrin formation. A qualitative comparison was then made between the experimental clot and the clot formed in silico. Although quantitative measurements were not made, there were similarities in the shapes and sizes of the clots. The validated computational model was used to compare clot formation under steady and pulsatile flow conditions. Realistic clot growth was observed and compared to the steady flow condition. It was found that a larger clot formed under pulsatile conditions. Clot formation with the presence of valve activity was also investigated. The effect of opening and closing of the valves was achieved by varying the blood flow diameter at the inlet instead of modelling the valves as solid walls and accounting for the leaflet movement by solving the governing equations for the fluid-solid interaction (FSI), as used in existing models. The model was then applied to a patient-specific geometry. Realistic clot growth was achieved using this model, and the clot was compared to a clot formed in vivo, as depicted in the original imaging scan. The model helps us better understand the clot growth process in the femoral vein on a patient-specific level. It also shows that the presence of venous valves increases the size of clot formed compared to steady flow. However, the high strain rate present makes the clot formed smaller than in standard pulsatile flow cases.
Open Access
Computational model of intraluminal thrombus growth in abdominal aortic aneurysms with fibrin generation
(2020) Taylor, Mark Robin; Ngoepe, Malebogo
Abdominal aortic aneurysms affect 0.2% of the population and are closely associated with intraluminal thromboses (ILTs) that develop in the sac. Advanced imaging and treatment techniques are available, however there is room for improvement in the methods used to predict the outcome or necessity of surgical intervention. For a computational model to be useful in this clinical setting, it would need to incorporate relevant patient-specific data and prioritise simplicity and speed over exhaustive detail. This paper presents the details of such a model, for abdominal aortic aneurysms, particularly in the simplification of the coagulation biochemistry. Explicit modelling of the coagulation cascade is replaced with a patient-specific thrombin generation curve. This curve is defined by three values obtained from a blood test. Another key feature is the thrombosis growth model, which incorporates conversion of fibrinogen to fibrin, variation between clot core and shell, and mechanical lysis. The model generates ILTs with morphologies visually similar to those typically found in the body, however more work is required to refine and validate the mode.
Open Access
Computational model of thrombosis in cerebral aneurysms for predicting clotting outcomes in flow diverter treated patient-derived geometries validated with novel PIV-based ln vitro clotting flow experiment
(2024) Hume, Struan Robertson; Ngoepe, Malebogo
There are a growing number of computational models of thrombosis in cerebral aneurysms designed with consideration towards clinical use and research. Many thrombosis models include complicated clotting mechanisms, which can be computationally expensive, and present a challenge to comprehensively validate in vitro due in part to the complexity of adequately measuring the ongoing interaction between flow and clot-growth; a key factor in predicting aneurysm-occlusion after surgical placement of a stent. To this end, a pulsatile-flow direct thrombosis-model has been developed towards use in a clinical environment to predict thrombosis outcomes in patient-specific cerebral aneurysm cases with and without a flow diverter, and is validated at each 0.05s timestep using a novel PIV-based (Particle Image Velocimetry) in vitro clotting flow experiment that simultaneously captures motion of a fibrin clot strand and surrounding flow within an idealized aneurysm flow vessel. The validated pulsatile-flow fibrin clot-model produces plausible clotting outcomes in each of the patient-specific cerebral aneurysm cases, with and without flow diverters, dependent upon the classification and size of cerebral aneurysm in question. The novel PIV-based in vitro clotting flow experiment demonstrates that fibrin clotting and flow may be measured simultaneously using PIV techniques. In cross-referencing the results of multiple simulations and flow experiments performed for this thesis with one another and to literature, the combined studies indicate two potentially important considerations for future direct thrombosis models of cerebral aneurysms. These include directional clot growth in accordance with the alignment of fibrin strands due to periodically high physiological flow rates, and the significance of the non-Newtonian features of blood for the modelling of physiological flow and wall boundaries in major cerebral arteries, although the results of a small sample of experiments is far from conclusive and further study in these areas is required.
Open Access
Computational modelling of hydrogel therapies
(2025) Ahmed, Sadman Sakib; Ngoepe, Malebogo; Ho, Wei Hua
Myocardial infarctions (heart attacks) are a type of cardiovascular disease that affects a large population of people around the world. They lead to the death of heart tissue, which is eventually replaced by scar tissue in a non-reversible process. Scar tissue does not behave like normal heart tissue (myocardium), and this leads to a decrease in heart function and eventually, heart failure. Current areas of research regarding treatment of this disease look at using injectable biomaterials to provide mechanical support to existing scar tissue. This has been shown to improve heart function in various animal models. A popular biomaterial of choice is polyethylene glycol (PEG), chosen for its biocompatibility and other desirable qualities. PEG undergoes a gelation process, where it changes from a liquid to a gel via a chemical reaction. This is useful as it can be injected during its liquid state and can then solidify into a gel, over a certain period, at a location of interest. Previous in situ studies have noted that the gel that is injected in the myocardium is found in other parts of the body. This is undesirable as this may lead to adverse side effects if the gel solidifies elsewhere in the body. PEG is relatively expensive, and it is also of interest to optimize the procedure to use enough of it. The hypothesis for the gel ending up elsewhere in the body is that the greatest losses of the gel occur while it is a liquid. This research aims to answer the hypothesis by developing a computational framework that investigates the flow behavior of PEG present in rat myocardium as it undergoes gelation. A methodology is presented for characterizing the gelation of PEG from existing rheology data. A material model is developed for gelation by using a time-dependent viscosity model that is implemented numerically in Ansys Polyflow. A second methodology is presented for modelling the flow of PEG out of a domain of interest using existing FSI results. This methodology utilizes a traction boundary condition, which, when applied to a domain of interest, results in outflows out of all orifices. 2D computational studies are carried out to characterize the impact of applied traction on observed flow rate. The studies are done across the range of viscosities for which the liquid gel exists and explore the use of a time-dependent viscosity. This is done using an idealized, microfluidic geometry that is derived from literature. The findings from the 2D study are used to build a 3D model that uses realistic geometry of PEG contained in rat myocardium. 3D computational studies are conducted to explore the aforementioned hypothesis. The findings from the studies show the gel exists at its lowest viscosity for a relatively long period of time, during which it incurs significant losses out of the myocardium. The findings also show that for an initial increase in viscosity due to gelation, the rate at which the losses occur decreases significantly. However, subsequent increases in viscosity do not result in an equal decrease in the rate of loss; i.e., as viscosity increases during gelation, the rate at which losses occur decreases slowly. The work presented can be used to support the development of PEG for future studies and gives insight into optimizing the procedure for injecting PEG into myocardium. Furthermore, the framework can be used to investigate the flow behaviour of PEG when injected into different parts of the heart.
Open Access
Design, verification and validation of a dynamic model for an intramuscular autoinjector
(2025) Magubane, Ntokozo; Sivarasu, Sudesh; Ngoepe, Malebogo
The rising prevalence of autoimmune and chronic conditions is a concern worldwide, leading to a need for disease management therapies to aid medication adherence and compliance. Autoinjectors are prime medical devices used to inject antidotes and prophylactics into the intramuscular or subcutaneous layer. These devices are used in various autoimmune and chronic conditions to counteract opioid overdose and nerve gas poisoning. There are easy to use and often allow for self-administration of medication. The development of new injectable molecules for different diseases and the reformation of first-generation pharmaceuticals has led to growing interest in making autoinjectors usable for different medications. The major challenge in this quest is the complexity of injecting highly viscous drugs and in large volumes. This research aims to develop a dynamic model to describe the influence of drug viscosity and volume on the injection process and evaluate the sensitivity of medication fluid behaviour to variations in component dimensions of the autoinjector fluid delivery system. Using mathematical modelling, the kinematic properties relating to the plunger motion were modelled and verified through physical testing. This model was optimised via the evaluation of sensitivity and measuring the results against the validation results. A Computational Fluid Dynamics (CFD) study was conducted to analyse the fluid behaviour of different viscous medications. This study allowed for the variation of fluid viscosity, medication volume, needle gauge and length. The computational model was then verified and validated using the American Standard of Mechanical Engineers Verification and Validation Standards (ASME V&V 40 and V&V20). This was accompanied by fluid characterisation of four medications, namely adrenaline, amikacin, Vaxigrip and insulin basaglar using a rheometer. The ZwickRoell universal tester measured the syringe force and plunger displacement to derive the pressure changes. These results were evaluated against the computational model. According to the optimised mathematical model and validation results, the plunger displacement increases linearly when the plunger motion is initiated until a maximum displacement is reached. Separation flow was observed in the syringe for viscosities between 15 - 80 cP. This represents decreasing flow as pressure increases in the syringe for medications with a viscosity in this range. This phenomenon increases the chances of needle deformation and injection pain due to tissue damage. High-gauge needles are more effective for injecting lower volumes and low viscosity medications, while low-gauge needles work best for injecting larger volumes or highly viscous medications. The model risk is defined to be high according to the ASME guidelines. If the model results directly impact the autoinjector design without any further testing to inform the design process, an incorrect model would result in poor-performing autoinjectors that fail to deliver the desired dose of medication into the intramuscular layer at the required injection time. This research has proved that it's possible to inject highly viscous drugs and large doses using an autoinjector if the right balance of injection force, injection time and needle dimensions is carefully selected to improve compliance and patient experience.
Open Access
Evaluating the impact of ultraviolet light on chemical and mechanical properties of human scalp hair through controlled and natural experiments
(2025) Buthelezi, Ntandoyenkosi; Ngoepe, Malebogo; Khumalo Nonhlanhla
Introduction: Human hair fibres have been studied in a variety of fields for different applications. Examples include the use of hair as a medical testing substrate for diagnosis of pathologies, as a tool for forensic investigations, a substrate for cosmetic product development and as a precursor material for construction engineering and agriculture. The changes that affect the character and behaviour of hair fibres, arising from exposure to environmental factors such as UV radiation, have been the focus of some studies. Even though there are a range of environmental variables that could be worthy of consideration, understanding the impact of UV radiation is increasingly important given the inevitability of exposure and changes in climatic conditions. Chemical and mechanical experiments have demonstrated that the proteins and pigments of human hair are most heavily affected during exposure to UV radiation, resulting in unhealthy fibres, suppressed growth and permanent hair loss. Problem: Photodamage and photoprotective potential of human hair fibres have been identified, however many studies have been conducted on ‘Caucasian' hair samples and the findings have been assumed to hold for other population groups. The use of race-based classification systems has further confounded findings, as the groupings are not able to account for intra-population and intra-individual variation objectively. This study is the first evaluation aimed at comparing the effect of UV radiation on hair of varying curl, using a combination of chemical and mechanical tools alongside advanced statistical techniques on the same samples. Furthermore, the study makes use of both controlled UV-exposed fibres and dreadlocks exposed to natural sunlight. Therefore, this study aims to compare the effects of UV irradiation and hair curl on the chemical and mechanical characteristics on human hair fibres, for controlled UV-exposed fibres and natural sun-bleached dreadlocks. Method: To analyse the influence of the radiation interaction with human hair, natural untreated, and UVA and UVB radiated black hair samples of low to high curl were compared. This study utilized thermogravimetric analysis and derivative thermogravimetry (TGA/DTG), Fourier-transform infrared spectrometry (FTIR- ATR) coupled with multivariate analysis, scanning electron microscopy (SEM) and mechanical tensile testing (MTT) techniques to obtain information on the chemical, and mechanical properties of the human hair. Results: UVB radiation resulted in more significant changes than UVA across all curl types, as demonstrated by all measurement techniques. After 7 days of UVB exposure, FTIR revealed changes on the absorbance of the amide and lipid bands. Major changes were observed associated with degradation of proteins at characteristic peaks; amide A (3300 and 3070 cm-1 ), amide I (1650 cm-1 ), and amide II (1550 cm-1 ). UVB 7-days exposure also produced the highest amount of the cystine oxidation products (at 1022-1077 cm-1 region) and had the lowest absorption for the disulphide bonds. The TGA/DTG showed that the onset degradation temperature (ODT) of UVB-treated fibres was significantly reduced compared to both the control group and UVA exposed group. The MTT results showed increased yield stress and yield strain for UVB exposed fibres. Interestingly, a threshold effect in damage was observed in low curl fibres as opposed to high curl fibres. Extending UV exposure time beyond 7-days in low curl hair fibres did not result in increased damage, however, this phenomenon was not observed for the high curl fibres. For the natural dreadlock experiment, the distal hair fibres showed reduced thermal stability and changes in the chemical properties observed on the FTIR. There were no significant changes on the mechanical properties of proximal compared to distal hair fibres. However, a SEM investigation revealed more structural damage in the distal fibres which might contribute to differences observed in the TGA and FTIR. This study provides novel insight into a critical point at which damage becomes pronounced and highlights how hair fibres of varying curl respond to UV-induced stress. This knowledge will not only aid hair scientists in navigating population specific hair analysis, but also provides valuable insights and paves a way for informed strategies for hair care development industry and related disciplines. The results obtained in this study could be beneficial for the hair care industry by informing the formulation of targeted products that account for different hair types and their specific vulnerabilities to UV damage. For instance, cosmetic companies could develop specialized UV protection shampoos and conditioners designed to address the specific needs to different hair types.
Open Access
Growth in a computational planning pipeline for treatment in patient-specific coarctation of the aorta
(2024) Hampwaye, Nasonkwe; Ngoepe, Malebogo
Coarctation of the Aorta (CoA) is a Congenital Heart Disease (CHD) that is present at birth and is usually detected in the early years of the child. In the individualized treatment of a CoA patient, a non-severe case which initially exhibits no symptoms, and thus no treatment is necessary, could potentially turn severe due to the growth of the baby. Growth is characteristic of living structures including the aorta and is understood to occur in response to a given mechanica
Restricted
Interdependence between geometric, tensile and chemical bond behaviours of untreated hair fibres
(2020) Cloete, Elsabe; Ngoepe, Malebogo; Khumalo, Nonhlanhla
To date, an accurate understanding of the dynamics between the fibre's inherent geometric, mechanical and biological characteristics is deficient, affecting the reliability and robustness of hair data. There is also insufficient scientific clarity on the behaviour of curly hair, as most of the conclusions have been drawn from studies focusing on straight fibres. This research project aimed at gaining a more accurate understanding of the interrelationships between fibre curliness, strength and chemical bonding. In the current understanding of hair mechanics, curly fibres are considered to have a lower tensile strength than straight fibres. Furthermore, the current understanding of hair fibres does not associate hydrogen bonding with fibre shape. During experimentation, inadvertent observations suggested that current tensile methods ignore an important component of hair strength in curly fibres, and that hydrogen bonding supports fibre curliness. Intensive scrutiny of these observations led to fundamental contributions to the understanding of curly hair. Research tools included tensile, geometric, image, (FTIR) spectroscopic assessments, regression modelling and multivariate statistical analysis. Through this research, the role of hydrogen bonding in fibre curliness has been established. A theory is presented about extraordinary hydrogen bonds and the existence of hydrogen bond networks across the fibre matrix of curly hair. The theory has been substantiated experimentally via FTIR and weight measurements. The research also established the importance of the preelastic tensile region for curly fibres. It was clearly demonstrated that tensile strength of hair fibres is not only dependent on Young's modulus, but also on the fibre's inherent viscoelasticity, which appears to be important in curly fibres but becomes negligible with loss of curl. A model, developed from experimental observations and insights from similar biological fibres, is also presented. The model gives insights into ultrastructural changes at the early onset of fibre elongation. It also demonstrates the association between viscoelasticity and hydrogen bond networks. Taking this into consideration, a constitutive equation, developed to determine hair fibre strength accurately, is also presented in this work. This work does not replace current fibre curvature theories, but provides additional insights into hair shape, and therefore presents a fundamental contribution to curvature in human hair. It also highlights the shortcomings of current instrumentation methods that contribute to inaccurate conclusions regarding the strength of curly fibres.
Open Access
Microstructural non-linear finite-element analysis of rat myocardium with hydrogel biomaterial inclusions
(2025) Manack, Uzair; Alheit, Benjamin; Ngoepe, Malebogo
Hydrogel biomaterial injectate therapies have emerged as a promising treatment modality for myocardial infarction (MI). Studies conducted on small and large animal models have yielded positive results in improving cardiac function and reducing adverse ventricular remodelling post-MI. These therapies have also recently entered phase I and II human clinical trials, with limited positive results, but no significant adverse effects. Computational modelling has been used extensively to investigate the potential effects of hydrogel injec tate therapies, due to the risk-free and repeatable nature of these tests. Macroscale cardiac computational models are used to investigate the full-scale behaviour of the heart, while microscale models yield infor mation on the behaviour of the cardiac microstructure. In order to reduce computational expense, many existing studies make use of idealisations regarding the macroscale or microscale cardiac geometry, as well as the physical behaviour of both the cardiac tissue and hydrogel injectate. The aim of the current study was to develop a computational framework that reduced the need for idealisations of the cardiac microstructural geometry, evaluated the validity of the assumption that both the cardiac tissue and hy drogel injectate could be described as elastic solids, and provided a basis for extension to more complex descriptions of material behaviour. A realistic microstructural finite-element (FE) mesh was reconstructed from high-resolution confocal mi croscopy imaging data of rat myocardium. The reconstructed mesh did not necessitate idealisations of the cardiac tissue structure or the distribution of the hydrogel injectate. To investigate the mechan ical response of the microstructure, under the assumption that both the cardiac tissue and hydrogel behaved as elastic solids, an FE solver was developed using the open-source FE library deal.II. The solver was capable of implementing both isotropic and anisotropic hyperelastic material models, and applying thermodynamically-admissible boundary conditions to the microstructure. Suitable boundary conditions were derived from the results of an existing macroscale FE model of rat myocardium, and used to investigate the mechanical response of the microstructure under five possible loading scenarios. The results indicated that, under certain loading conditions, the observed stresses in the microstructure significantly exceeded reasonable elastic limits for the materials. This provides an indication that the assumption of elastic material behaviour is not always suitable when conducting in silico investigations of cardiac tissue and hydrogel injectate, and serves as a justification for the use of alternative descriptions of material behaviour. Furthermore, the framework was shown to be capable of implementing both static and time-dependent boundary conditions. This functionality provides the basis for the framework to be extended to more advanced models such as viscoelasticity and poroelasticity, which have been implemented in other studies using the deal.II library
Open Access
Numerical optimisation and theoretical analysis of complex microchannel heat exchangers
(2025) Godi, Nahum Yustus; Collier-Reed, Brandon; Ngoepe, Malebogo
The purpose of this study is to evaluate the performance of the geometries in forced convective heat transfer and steady state laminar incompressible fluid flow. Microchannel heat sinks used were numerically modelled from highly conductive (aluminium) solid material substrate. ANSYS FLUENT Response Surface Optimisation Tool (RSO) was used to numerically optimise and compare the performance of combined microchannel heat sinks with perforated, solid, half-hollow and hollow fins. The simulation began by optimising a typical microchannel heat sink geometry before fin-bars were inserted into the cooling channel to augment heat transfer. Furthermore, solid and perforated fins were modelled and added on the top of the typical heat sink. The performance of the various configurations was then compared. A novel combined microchannel design with circular micro fins was modelled with a circular flow channel. Additionally, a hybrid model was developed, incorporating circular fins on a microchannel heat sink with a rectangular flow channel. The third design featured rectangular fins mounted on a microchannel with a rectangular flow channel. The combined microchannels, featuring circular and rectangular fins, were cooled by water flowing through the channels and internally along the fin inner surface walls to dissipate heat. These designs were then integrated into the computational domain and subjected to cooling using both water and an air stream. The water flows through the flow channel while air flows over the vertical fins to remove excess heat from the external wall surfaces of the fins in forced convection laminar flow condition. Theoretical analysis (intersection of asymptotes method) was carried out in the cooling channels (circular and rectangular). The theoretical analysis results indicated the presence of an optimal geometry among the various cross-sectional shapes, effectively cooling a volume with a uniformly distributed heat flux. A comparison of the analytic findings with the numerical results demonstrates that an optimal design is possible. The numerical cooling processes were carried out in parallel and counter flows. These findings demonstrate that an optimal design can be realised with a combination of Computational Fluid Dynamics and geometric modelling techniques.
Open Access
Pulsatile Flow in Computational Modelling of Thrombosis in Cerebral Aneurysms
(2019) Hume, Struan; Ngoepe, Malebogo; Ho, Wei Hua
Ngoepe and Ventikos have developed one of a growing number of computational models of thrombosis of cerebral aneurysms designed with consideration towards clinical use and research. Their model, amongst many others, utilizes computationally inexpensive steady flow conditions. However, pulsatile flow better characterizes blood flow in-vivo. Steady flow is an acceptable approximation of pulsatile flow from a fluid dynamics perspective, but there is no prior evidence suggesting whether it is an acceptable approximation when considering clot formation within a flowing environment. To this end a pulsatile flow model has been created in ANSYS® Fluent, and a function from Ngoepe and Ventikos’s computational model that simulates the release of thrombin, a chemical responsible for clotting activation, has been implemented. The output of this simulation is compared to the output of an otherwise identical simulation utilizing Particle-Image-Velocimetry (PIV) validated steady flow conditions, to determine whether clotting outcome of Ngoepe and Ventikos’s model, amongst others, differs with pulsatile flow This experiment revealed that the concentration of thrombin required for clotting activation is generated in nearly half the time when utilizing pulsatile flow over steady flow. Pulsatile flow creates unsteady flow patterns within the aneurysm, which create an environment where less thrombin is carried out of the aneurysm and into the regular bloodstream. This indicates that steady flow approximations for realistic clotting in computational models of thrombosis of cerebral aneurysms without strong consideration for the effects of pulsatile flow are inaccurate.
Open Access
The Development of a Patient-Specific, Open Source Computational Fluid Dynamics Tool to Comprehensively and Innovatively Study Coarctation of the Aorta in a Limited Resource Clinical Context
(2020) Swanson, Liam; Ngoepe, Malebogo; Zuhlke, Liesl
Congenital heart disease (CHD) has a global prevalence of 8 per 1000 births [1] and coarctation of the aorta (CoA) is one of the most common defects with a prevalence of 7% of all cases. The occurrence of CHD in Africa is estimated to be significantly lower, which is attributed to a lack of data [2]. This emphasises the restricted human resources, as well as diagnostic and intervention capacity of specialists in Africa which leads to delayed treatment, presentation with established severity and, consequently, a worse prognosis. Computational Fluid Dynamics (CFD) is seen as the tool that will lead to a better understanding of the haemodynamic effects caused by the malformations related to CoA and provide insights into post-repair morbidity. In addition, the development of a computational tool is envisaged to improve the clinical capacity for diagnosis as well as provide a tool to conduct in silico repair planning. In a low and lower-middle income country healthcare facility, the supplementary data that CFD can provide can add diagnostic value, plan interventions to be more effective and efficient, as well as provide data that may improve postrepair patient management. The aim of this project is to develop a patient-specific, open source, computational fluid dynamics toolchain that is able to study the haemodynamics relating to CoA. In order to do so, a protocol for the collection of doppler echocardiography (echo) and CTA data is proposed. The method for processing the echo data and manually segmenting the CTA data is presented and evaluated. The open source, OpenFOAM code is used to simulate a patient-specific CoA case as well as two in silico designs of coarctation repairs based on expanding the coarctation from the original dataset. The CFD toolchain was developed such that patient data collected from the hospital could be processed to present key haemodynamic metrics such as velocities in the field at the coarctation zone, the pressure gradient across the coarctation and volumetric flow rates through each supra-aortic branch. These results are obtained for each case's geometry, and the trends and impacts that increasing the coarctation ratio has on each of the haemodynamic metrics is presented. The results show that the coarctation pressure gradient and maximum coarctation velocity decrease while perfusion of the lower limbs recovers with expanding coarctation ratio. Following an analysis of the results, it is evident that the pipeline is capable of running patient-specific CFD simulations and can present clinically relevant results. It is noted that this work is a proof of concept and so several steps are discussed that will improve the pipeline.
Open Access
The Development of a Reduced Order Model for Prediction of Haemodynamic and Biochemical Changes in a Computational Cerebral Aneurysm Thrombosis Model
(2023) Ngwenya, Tinashe; Ngoepe, Malebogo
A cerebral aneurysm is a pathological, localized, and irreversible cerebral arterial vessel wall dilation. Cerebral aneurysms pose a risk of subarachnoid haemorrhage if they rupture. A ruptured cerebral aneurysm could lead to mortality or morbidity. It is therefore important for clinicians to be able to predict whether an aneurysm will rupture or not. This enables them to decide whether to perform surgical or endovascular procedures or not intervene at all. However, medical imaging is not always possible, thus there is a need for the development of computational tools to assist with predictions. A computational fluid dynamics (CFD) model that couples haemodynamics and biochemistry was developed to simulate cerebral aneurysm thrombosis. The blood was modelled as a Newtonian fluid and thrombin generation curves (TGCs) were used to model the release of thrombin from the injured vessel walls. TGCs representative of haemophiliac, healthy and thrombotic patients were used. Thrombin reacts with fibrinogen to produce fibrin and form a clot within the aneurysmal sac. The CFD results were used to build reduced-order models using machine learning algorithms. Multiple polynomial regression and logistic regression machine learning algorithms were used to predict clot size in patients. The K-nearest neighbours algorithm was used to develop a model that classifies patients' clotting profiles. The biochemistry was found to be more sensitive to mesh size compared to the haemodynamics. Large timesteps overpredicted clot size in pulsatile flow. The rate of clot growth in pulsatile and plug flow were different; the predicted clot size in pulsatile flow was 14.6% greater than in plug flow. When variable diffusivity was used, the predicted clot size was 25.4% less than that with constant diffusivity. The numerical model was validated against the experimental results of Ngoepe et al. and there was good agreement in the predictions. Vortical structures that formed in the aneurysm sac evolved differently in the haemophiliac, healthy and thrombotic cases. The clot size, biochemistry and haemodynamics were found to be interdependent. Only the thrombotic case had full occlusion in the 5000 seconds of simulation time considered, however, the haemophiliac and healthy case had more than 90% occlusion. The clot in the healthy case was larger than in the haemophiliac patient. iv Logistic regression fitted the CFD data better than multiple polynomial regression. The KNN algorithm produced satisfactory decision boundaries and classified the patients effectively. The reduced order models developed from the classification and regression algorithms could assist clinicians in interventional planning and reduce the cost of health care.

Browsing by Author "Ngoepe, Malebogo"

Results Per Page

Sort Options