Harshinee Goordoyal

Machine Learning for Predictions of Haemodynamics in Cardiovascular Devices

Project Summary

Cardiovascular diseases are the leading cause of death globally with 18 million deaths/year. Many cardiovascular diseases require treatment with blood contacting medical devices such as vascular stents, prosthetic valves and even artificial hearts. Computational Fluid Dynamics (CFD) is used in the design of these cardiovascular devices. However, blood flow is complicated by the non-Newtonian rheology of blood, the pulsatile and transitional flow characteristics, and the presence of cells and proteins. This complexity limits the accuracy of calculations, which are also time consuming and expensive. Inaccurate calculations lead to poor design choices, misunderstanding of the fundamental science, and lack of trust in simulation results. Improved methods for simulating both the fluid dynamics and biological interactions are required for CFD to fulfil its potential within the field.

This project will explore the use of artificial intelligence (AI) in CFD simulations of blood flow in artificial hearts, and in simpler geometries giving flow features characteristic of those found in cardiovascular devices. The project will focus on one or more of the following main ideas: the use of Fourier inspired methods to learn from oscillating flows with single frequencies in order to predict physiological pulsatile flows through rotating domains; refinement of the Reynolds Stress tensor by learning from Large Eddy Simulations (LES) to improve the use of unsteady Reynolds Averaged Navier Stokes (URANS) modelling in transitionally turbulent flows; use of data from turbulent eddies gained from LES for creating numerical models for haemolysis (damage to the red blood cells) implemented in URANS. Each of the ideas treats fluid dynamic simulations in a multiscale way, with the aim of achieving the accuracy of a high fidelity simulation, by using only a coarse simulation augmented with supplementary models learned from high fidelity simulations.

Research Interests

Although I am a mechanical engineer by training, I have always been inclined towards the application of my engineering knowledge in a medical context.

My determination and dedication to pursue research in the field was further reinforced as two close members of my family were recently diagnosed with cardiovascular disease.

I aim to research ways of improving the design of cardiovascular and medical devices to help elevate the quality and length of life of cardiovascular disease patients.

Background

Mechanical Engineering (MEng), University of Bristol, UK.

Supervisors

Dr Katharine Fraser