| Atherosclerotic plaques frequently extend into the vessel lumen, decreasing blood flow to distal tissues. Accurate assessment of the severity of the resulting luminal stenoses is critical to clinical management of these lesions. Computed tomography (CT) angiography is an important diagnostic technique used to non-invasively assess vascular anatomy in 3 dimensions. Both the iodine-filled lumen and the impinging plaque can be well visualized. The presence of calcium in these plaques, however, can cause errors in percent area luminal stenosis estimates. This is because the signal from dense calcium can have the same intensity as the iodinated lumen, making stenosis quantification highly subjective and error prone. Additionally, the bright signal from the dense calcium alters the signal of the adjacent iodinated lumen due to limitations in the spatial resolution of the system; this effect is referred to as calcium blooming, and describes the contamination of the iodine signal with calcium signal due to the system blurring (i.e. point spread function). For both of these reasons, a technique to differentiate the iodine and calcium signals from each other will increase the accuracy and reproducibility of non-invasive estimation of stenosis severity using CT angiography.;The body of work performed throughout this project uses a novel type of CT x-ray detector, known as a photon-counting detector, which can resolve the energies of detected photons individually. By binning the detected photons according to their energies, one can identify elements of different atomic numbers, in this case calcium and iodine. This process is referred to as material decomposition, as it "decomposes" the signal into the base materials present in the object. To date we have developed material decomposition techniques that separate water-like materials (e.g. tissue) from iodine and calcium, as these are the three most critical base materials involved in the quantitation of stenosis severity. Material decomposition techniques have been implemented using two x-ray spectra, a strategy referred to as dual-energy CT. However, assumptions regarding volume conservation are required which do not always hold, particularly when iodine is mixed with blood. To avoid this, more than 2 energy measurements are required. To accomplish this, a research whole-body photon-counting CT (PCCT) scanner (SOMATOM CounT, Siemens Healthcare, Forchheim, Germany) was used in this study.;This thesis work was separated into 4 distinct aims, which have all been accomplished. Aim 1: Knowledge of the physics interactions in the photon counting detector was combined with knowledge of the scanner characteristics, data acquisition processes, data correction, and reconstruction methods to create a software simulation package to predict the signal, noise, and images of the actual PCCT scanner. Aim 2: A previously developed adaptive, non-local-means noise reduction method was adapted for use with the PCCT system to take advantage of data redundancies in the energy and spatial domains, and validated with experimental data. Aim 3: A 3-material decomposition algorithm was developed and validated using experimental data. The unique feature of the method is that, unlike commercial dual-energy CT (DECT), it does not require volume conservation of the materials in the mixture. It also does not require accurate information about x-ray spectra and detector response, making it much easier to implement without the assistance of the manufacturer. Aim 4: Phantom experiments were performed to demonstrate the accuracy of the proposed technique for estimating percent area luminal stenosis in the presence of dense calcifications.;Phantom experiments demonstrated that the developed software simulation package could accurately estimate the signal, noise, and images of the physical PCCT scanner (Aim 1). Multi-energy Non-local Means (MENLM) achieved about 80% noise reduction without affecting spatial or spectral resolution, while improving the low-contrast detection (Aim 2). The developed image-domain material decomposition method showed accurate estimation of base material densities, with root-mean-square-error of 4.3, 0.7, and 5.7 mg/ml (percent error: 6.20%, 7.18% and 0.60%) for calcium, iodine, and water, respectively, at clinical dose levels. In addition, material decomposition results in data from living swine demonstrated good separation of iodine, calcium, and water (Aim 3). Finally, experimental validation of the hypothesis was performed in Aim 4, where the estimation accuracy was evaluated by phantoms with different degrees of stenoses, vessel diameters, and calcification densities. For stenosis quantitation, a maximum estimation error of 10% was found for PCCT---almost half of what was found for DECT (18%)---using phantoms with known lumen and calcium dimensions (Aim 4). These errors were lower than what was achieved using current clinical techniques, which involve single-energy CT and threshold-based segmentation of the stenotic vessels. Additionally, feasibility of the developed material decomposition technique in actual carotid vessels has further been demonstrated, providing confidence that the accuracy in vivo will be similar to in phantoms.;These results support the hypothesis that material decomposition of multi-energy CT images can be used to accurately differentiate iodine from calcium, which can be exploited to accurately and reproducibly estimate stenosis severity by CT angiography. |
