Thesis

Compact diffuse correlation spectroscopy systems

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2024
Thesis identifier
  • T16929
Person Identifier (Local)
  • 202061057
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • Diffuse correlation spectroscopy (DCS) is a powerful tool for investigating microvascular dynamics in deep tissues. It has been used for non-invasive blood flow assessment at the bedside. This study first provides a thorough literature review on system setups (continuous-wave, frequency-domain, and time-domain) and derive corresponding theoretical models. I then present an innovative deep learning algorithm, DCS-NET, which is easy and robust to train, fast, and insensitive to measurement noise for data processing. The absolute blood flow index (BFi) at different depths with/without measurement noise was calculated, followed by a relative blood flow calculation. I then calculated the intrinsic sensitivity, and calculated BFi with varied optical properties and scalp/skull thicknesses finally. Compared with the semi-infinite, and three-layer fitting methods, I show that the DCS-NET is approximately 17,000 times faster than the traditional three-layer model and 32 times faster than the semi-infinite model, respectively. It provides increased inherent sensitivity to deep tissues compared to fitting methods. DCS-NET demonstrates remarkable noise resilience and is minimally affected by variations in μa and μs'. Additionally, we have shown that DCS-NET can extract relative blood flow index (rBFi) with a substantially lower error of 8.35%. In comparison, the semi-infinite and three-layer fitting models produce considerable errors in rBFi, amounting to 43.76% and 19.66%, respectively. Additionally, a DCS prototype is developed by integrating an advanced CMOS single-photon avalanche diode array, which employing a parallel light detection scheme, exhibits exceptional photon-counting throughput. The system tested on a milk phantom, showing SNR gain with the entire sensor is improved nearly 160-fold compared with a single pixel. An in vivo blood occlusion test was also performed. In conclusion, our system works well, and this research can offer peers effective guidance to embark on DCS research.
Advisor / supervisor
  • Chen, Yu, 1973 December 28-
  • Li, David Day-Uei
Resource Type
Note
  • This thesis was previously held under moratorium from 17th May 2024 until 17th may 2025.
DOI

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T16929 2024-05-31 Public Download