| dc.description.abstract | Intravascular imaging is the gold standard imaging modality for assessing and characterising plaques. However, it’s invasive with small risk of complications so cannot be adopted as a risk stratification tool for all patients. Coronary computed tomography angiography (CCTA) is a non-invasive imaging modality with good diagnostic accuracy for detecting coronary artery disease. However, CCTA’s ability to quantify and characterise plaque remains inferior to intravascular imaging. This thesis is designed to explore and enhance the performance of CCTA in assessing plaque pathology using state of the art intravascular imaging as the reference standard. In the first part of this thesis, we assessed the potential of optical coherence tomography (OCT) in assessing plaque burden using a novel post-processing methodology, the attenuation-compensation technique which allowed improved visualisation of the external elastic membrane and quantification of plaque burden. This technique was superior to conventional OCT for detecting the outer vessel wall border but was less capable of detecting and quantifying lipid and calcific tissue as well as plaque micro-characteristics. In view of the limited efficacy of OCT to provide a complete assessment of plaque pathology, for the second part of this thesis, to study the efficacy and improve the efficacy of CCTA, we used infrared-spectroscopy – intravascular imaging (NIRS-IVUS) as the reference standard. This modality is the only FDA approved modality for assessing vulnerable (lipid-rich) plaques. Findings presented herein demonstrate that CCTA images reconstructed with thinner slice thickness (0.50mm) and highest strength model-based iterative reconstruction (ADMIRE 5) provided closest estimations to NIRS-IVUS. In addition, although CCTA and NIRS-IVUS detected similar number of plaques seen on coronary angiography, CCTA detected two-thirds of all plaques seen on NIRS-IVUS, highlighting its limitation in qualitative and quantitative plaque assessment. In order to improve the performance of CCTA, a deep-learning methodology was trained using NIRS-IVUS annotations, to provide fast and closer estimations to NIRS-IVUS for detecting and quantifying lesions and assessing plaque features associated with increased vulnerability. Testing of this method using NIRS-IVUS annotations showed excellent results. The final part of the thesis focused on the accurate computation of the distribution of the local haemodynamic forces that are well established instigators of plaque progression and compared two reconstruction methodologies developed for the fusion of intravascular imaging and angiography. We found that the methodology that incorporates vessel side branches in 3D model geometry enables more accurate shear stress computation and prediction of plaque progression compared to the methodology that was focused solely on the modeling of the main branch. In conclusion, this thesis highlights the limitations of current CCTA analysis and offers a deep-learning solution, which in combination with accurate assessment of local haemodynamic forces may allow better cardiovascular risk stratification and prediction of vulnerable plaques that are likely to progress and cause future cardiovascular events. | en_US |
