Computer Simulation of a Nitric Oxide-Releasing Catheter with a Novel Stable Convection-Diffusion Equation Solver and Automatic Quantification of Lung Ultrasound Comets by Machine Learning

Abstract

Biological transport processes often involve a boundary acting as separation of flow, most commonly in transport involving blood-contacting medical devices. The separation of flow creates two different scenarios of mass transport across the interface. No flow exists within the medical device and diffusion governs mass transport; both convection and diffusion exist when flow is present. The added convection creates a large concentration gradient around the interface. Computer simulation of such cases prove to be difficult and require proper shock capturing methods for the solutions to be stable, which is typically lacking in commercial solvers. In this thesis, we propose a second-order accurate numerical method for solving the convection-diffusion equation by using a gradient-limited Godunov-type convective flux and the multi-point flux approximation (MPFA) L-Method for the diffusion flux. We applied our solver towards simulation of a nitric oxide-releasing intravascular catheter. Intravascular catheters are essential for long-term vascular access in both diagnosis and treatment. Use of catheters are associated with risks for infection and thrombosis. Because infection and thrombosis lead to impaired flow and potentiality life threatening systemic infections, this leads to increased morbidity and mortality, requiring catheters to be replaced among other treatments for these complications. Nitric oxide (NO) is a potent antimicrobial and antithrombotic agent produced by vascular endothelial cells. The production level in vivo is so low that the physiological effects can only be seen around the endothelial cells. The catheter can incorporate a NO source in two major ways: by impregnating the catheter with NO-releasing compounds such as S-nitroso-N-acetyl penicillamine (SNAP) or using electrochemical reactions to generate NO from nitrites. We applied our solver to both situations to guide the design of the catheter. Simulations revealed that dissolved NO inside the catheter is depleted after 12 minutes without resupplying, and electrochemical release of NO requires 10.5 minutes to reach steady state. Lung edema is often present in patients with end-stage renal disease due to reduced filtration functions of the kidney. These patients require regular dialysis sessions to manage their fluid status. The clinical gold standard to quantify lung edema is to use CT, which exposes patients to high amounts of radiation and is not cost efficient. Fluid management in such patients becomes very challenging without a clear guideline of fluid to be removed during dialysis sessions. Hypotension during dialysis can limit fluid removal, even in the setting of ongoing fluid overload or congestive heart failure. Accurate assessment of the pulmonary fluid status is needed, so that fluid overload and congestive heart failure can be detected, especially in the setting of hypotension, allowing dialysis to be altered to improve fluid removal. Recently, reverberations in ultrasound signals, referred to as ``lung comets'' have emerged as a potential quantitative way to measure lung edema. Increased presence of lung comets is associated with higher amounts of pulmonary edema, higher mortality, and more adverse cardiac events. However, the lung comets are often counted by hand by physicians with single frames in lung ultrasound and high subjectivity has been found to exist among the counting by physicians. We applied image processing and neural network techniques as an attempt to provide an objective and accurate measurement of the amount of lung comets present. Our quantitative results are significantly correlated with diastolic blood pressure and ejection fraction.