Accurate measurement of blood flow in deep tissue is essential for diagnosing and monitoring a wide range of clinical conditions, yet current imaging methods are often invasive, limited to snapshots, or impractical for continuous bedside use. Diffuse correlation spectroscopy (DCS) offers a non-invasive optical alternative, measuring blood flow dynamics through temporal fluctuations in scattered light. However, modelling DCS signals remains computationally challenging, particularly when accounting for structured flow and tissue complexity.
This thesis investigates a coupled simulation framework that integrates particle dynamics from COMSOL with optical propagation computed via a modified Born series solver (WaveSim). The framework was used to test three validation cases: temperature-dependent Brownian motion, laminar flow, and parameter extraction through exponential fitting of the autocorrelation function.
The results demonstrate that the framework captures the qualitative behaviour of DCS. Brownian motion produced linear mean squared displacement (MSD) curves, with faster optical decorrelation at higher temperatures, while laminar flow introduced quadratic MSD growth and accelerated decay. However, quantitative agreement with theoretical scaling was not achieved. A key limitation was identified as a pixel-quantisation effect: particle displacements per frame were typically far smaller than the 0.2 μm WaveSim pixel size, rendering most motion invisible to the solver. This suppressed decorrelation, distorted temperature and flow scaling, and prevented unbiased parameter extraction. Additional constraints included limited statistical averaging, absence of phase tracking, reliance on decay rates without correlation diffusion equation (CDE) modelling, and high computational costs.
The findings confirm that the COMSOL–Born framework provides proof of concept for simulating DCS but requires significant refinement for quantitative accuracy. Future work should focus on subpixel rendering, improved temporal sampling, integration of CDE solutions, and domain decomposition. With these improvements, the framework has the potential to advance real-time, non-invasive monitoring of blood flow in clinical settings. i