Novel computational biomechanics for the assessment of coronary plaque vulnerability and its clinical implications
Huang, Jiayue
Huang, Jiayue
Loading...
Identifiers
http://hdl.handle.net/10379/17792
https://doi.org/10.13025/17367
https://doi.org/10.13025/17367
Repository DOI
Publication Date
2023-06-12
Type
Thesis
Downloads
Citation
Abstract
Acute coronary syndromes (ACS) continue to be one of the major causes of mortality around the world. Accumulating evidence indicates that thrombosis plays a critical role in the pathogenesis of ACS. Autopsy studies have revealed that plaque rupture remained the main cause of fatal coronary thrombosis, highlighting the need for early detection of plaques that are prone to rupture. Coronary biomechanical assessment provides a promising tool for the detection of vulnerable plaque, as plaque rupture is ultimately a mechanical process that can be accurately modelled, simulated and predicted from the interplay between flow, pressures and plaque composition within the coronary wall. The objective of this thesis is to develop novel methods for the biomechanical assessment of coronary plaque vulnerability. Finite element analysis (FEA) provides a valuable solution for the assessment of biomechanical environment within the vessel wall. One of the prerequisites for precise FEA results is the availability of an accurate geometric model. A fully automated deep learning (DL)-based plaque characterization algorithm was recently developed for optical coherence tomography (OCT) images, showing the potential for providing detailed geometric models for biomechanical assessment. Chapter 2 of this thesis validated the reliability of this DL algorithm in detecting calcified plaque, the component which plays an essential role for determining biomechanical environment within the vessel wall. The validation for OCT-DL was achieved using comprehensive tissue characterization technologies including OCT-derived optical properties, intravascular ultrasound (IVUS)-VH and echogenicity on accurately aligned in-vivo imaging of human coronary arteries. The concordance between OCT-DL and other modalities was assessed by kappa statistics. As a result, forty-three calcified plaques were detected by DL in 72 matched anatomic slices, 41 (95%) were confirmed by optical properties. Weighted kappa between OCT-DL and GS-IVUS, IVUS-VH and echogenicity were 0.69, 0.60 and 0.60, respectively. After having excluded artefactual optical shadowing (n=5) generated by guidewire or platinum marker, kappa increased to 0.77, 0.68 and 0.69, with agreement ranging between 90% and 93%. OCT empowered by DL showed substantial agreement with optical and ultrasound signals, revealing the reliability of using DL-aided OCT to assess the biomechanical environment within the vessel wall. At the same time, accurate delineation of coronary artery outer boundary, specifically the vessel contour, ensures the reliable reconstruction of the arterial geometric model, which in turn allows for a precise assessment of coronary biomechanics. The good penetration of ultrasound signals through blood and soft tissue makes IVUS the standard modality for precise vessel contour delineation and plaque burden (PB) assessment. In contrast, the near-infrared light used for OCT limits its penetration depth to only 1.5–2.0 mm, potentially hindering the accurate vessel contour delineation using OCT, despite its high image resolution. OCT-DL incorporates the information from adjacent OCT crosssectional images and the prior knowledge of regular shape of vessel contour (either circular or elliptical). As a result, OCT-DL is able to extrapolate the vessel contour and to provide an estimation of the PB from OCT images. This advancement offers a more efficient and automated solution for reconstructing the artery geometric model. Chapter 3 of this thesis validated the performance of OCT-DL in vessel contour delineation and PB assessment using precisely co-localized IVUS images. OCT cross-sections were further divided into four subgroups with different media visibility level to investigate the impact of media visibility on numerical differences between OCT-derived and IVUS-derived PB. As a result, 64 paired OCT and IVUS cross-sections were compared. OCT-DL showed good concordance with IVUS for PB assessment (ICC=0.81, difference=-3.53 ± 6.17%, p<0.001). The numerical difference between OCT-DL-derived PB and IVUS-derived PB was not substantially impacted by missing segments of media visualization (p=0.21). OCT-DL showed a diagnostic accuracy of 92% in identifying PB>65%. This study proved the capability of OCT-DL in accurately estimating PB and delineating vessel contours, regardless of media visibility. Based on the precise geometric model reconstructed using OCT-DL, Chapter 4 of this thesis proposed a novel approach to derive the changes in plaque structural stress (ΔPSS) during the cardiac cycle in-vivo using a combination of OCT images and intracoronary pressure recordings. The study enrolled all the intermediate lesions from a previous OCT study. OCT cross-sections at representative positions within each lesion were selected for ΔPSS analysis. As a result, a total of 50 lesions from 41 vessels were analysed. A significant ΔPSS gradient was observed across the plaque, being maximal at the proximal shoulder (45.7 [32.3, 78.6] kPa), intermediate at minimal lumen area (MLA) (39.0 [30.8, 69.1] kPa) and minimal at the distal shoulder (35.1 [28.2, 72.3] kPa; p=0.046). The presence of lipidic plaques was observed in 82% of the diseased segments. Lipidic plaque area showed good correlation with relative lumen deformation (r=0.88, p<0.001). Larger relative lumen deformation and ΔPSS were observed in diseased segments, compared with normal segments (percent diameter change: 8.2 ± 4.2% versus 6.3 ± 2.3%, p=0.04; ΔPSS: 59.3 ± 48.2 kPa versus 27.5 ± 8.2 kPa, p<0.001). ΔPSS was positively correlated with plaque burden (r=0.37, p<0.001) and negatively correlated with fibrous cap thickness (r=-0.25, p=0.004). ΔPSS has been proven to be able to provide a feasible method for assessing plaque biomechanics in-vivo from OCT images, consistent with previous biomechanical and clinical studies based on different methodologies. Larger ΔPSS at proximal shoulder and MLA indicates the critical sites for future biomechanical assessment. Despite the precise biomechanical assessment from high-resolution OCT images, the penetration rate of intracoronary imaging during diagnostic coronary angiography is relatively low, particularly in underdeveloped regions. The development of a simplified and cost-effective alternative for biomechanical assessment remains an outstanding goal in the field. As the corner stone for the diagnosis and management of coronary artery disease, dynamic angiogram records directly how vessel moves and deforms over the entire cardiac cycle based on the imprint left on the lumen contour. Aided by artificial intelligence, a novel method has been proposed enabling real-time assessment of radial wall strain (RWS), which is computed as the relative luminal deformation throughout cardiac cycle, from coronary angiography (RWSAngio). Chapter 5 in this thesis validated the agreement between novel RWSAngio and RWS derived from OCT, using the methodology proposed in Chapter 4 as the established reference standard (RWSOCT). The study enrolled all lesions from a previous OCT study and OCT was automatically coregistered with angiography. RWSOCT was analysed using FEA on OCT cross-sections at 1- mm interval. The luminal deformation in the direction of minimal lumen diameter was used to derive RWSOCT, following same definition as RWSAngio. The maximal RWSOCT and RWSAngio at normal segments adjacent to interrogated lesion were also analysed. As a result, RWSOCT analysis was performed in 578 OCT cross-sections from 45 lesions stemming from 36 patients. RWSAngio showed good correlation and agreement with RWSOCT (r=0.91, p<0.001; Lin’s coefficient=0.85). RWSAngio in atherosclerotic segments was significantly higher than normal segments (12.6% [11.0, 16.0] versus 4.5% [2.9, 5.5], p<0.001). The intraclass correlation coefficient for intra- and inter-observer variability in repeated RWSAngio analysis were 0.92 (95% CI:0.87-0.95) and 0.88 (95% CI:0.81-0.92), respectively. The mean analysis time of RWSOCT and RWSAngio for each lesion was 95.0 ± 41.1 and 0.9 ± 0.1 minutes, respectively. This study demonstrated the reliability of biomechanical assessment solely from angiography. In conclusion, this thesis conducted comprehensive validation of the performance of OCT-DL algorithms, including the automated detection of calcified plaques, delineation of vessel contour and evaluation of plaque burden. These validations led to the development of a novel approach that utilizes the geometric model reconstructed from OCT-DL results for precise OCT-based biomechanical assessment. Additionally, this thesis proposed a simplified method that enables real-time biomechanical assessment solely from coronary angiography and validated this method against the FEA results from co-registered OCT. The excellent concordance that we have observed demonstrates great potential for using the routinely available coronary angiography as a cost-effective tool in order to assess biomechanical properties of diseased coronary arteries in large populations.
Funder
Publisher
NUI Galway