# Sensitivity analysis of factors in a microfluidics CFD model of coagulation and cardiac applications

## TL;DR

This study uses computational models to analyze how blood clots form in microfluidic systems and in the heart, aiming to improve clinical predictions of thrombosis.

## Contribution

The novel use of Gaussian Process Emulators enhances computational efficiency in simulating coagulation and thrombus formation.

## Key findings

- The model accurately captures fibrin formation in microchannels and identifies key factors influencing thrombus growth.
- The global sensitivity analysis reveals the relative influence of input parameters on clot formation dynamics.
- The model successfully simulates thrombogenesis in an idealized left atrium, showing potential for cardiac applications.

## Abstract

Coagulation is essential for haemostasis but can lead to harmful thrombus formation in conditions such as atrial fibrillation. Computational fluid dynamics (CFD) models that incorporate coagulation with blood flow can simulate this process, but their complexity often limits their use in clinical settings. This study focuses on fibrin formation during the peak thrombin phase, a brief but critical period in the thrombogram, and employs Gaussian Process Emulators to improve computational efficiency. A simplified coagulation model is integrated into a CFD framework and validated using data from an ex vivo experiment. Model inputs are varied within physiological ranges to train an emulator that predicts fibrin concentration and haemodynamic changes associated with thrombus development. A global sensitivity analysis (GSA) is performed to identify the relative influence of each input parameter. The model is then applied to a two-dimensional idealised representation of the left atrium (LA) to evaluate its suitability for cardiac simulations and to compare thrombus formation dynamics between small vessel and atrial flow. The model accurately captures fibrin formation in microchannels and the GSA and reveals potential mechanisms underlying thrombus growth in vessels while the LA simulation simulated various stages of thrombogenesis in the LA. The use of emulators enables efficient and precise predictions, enhancing the clinical feasibility of thrombosis modelling. These findings provide a foundation for the development of predictive tools to assess thrombus formation and stroke risk in patients.