High Quality Low-dose Myocardial CT Perfusion Imaging Method

Coronary artery disease (CAD) is a kind of myocardial damage which is caused by coronary artery atherosclerosis who induced stenosis or occlusion; it gives raise to the imbalance between coronary blood flow and myocardial oxygen demand. In 2012, the American Heart Association (AHA) reported that there are more than 1 million CAD attacks per year in the USA; China’s CAD incidence shows rapid growth trend and has gradually become a leading disease to people’s health and cause of death in China; according to the World Health Organization (WHO)’s report in 2011, the CAD induced deaths had been listed as the world’s second in China. Early accurate diagnosis of CAD is important for clinical treatment, evaluation of curative effect and prognosis; and this will lead to a great drop in the incidence of the cardiovascular diseases and the mortality from such diseases.Computerized tomography coronary angiography (CTCA) is the major diagnostic method for CAD at present, however, CTCA may be unclear to find the functional significance of a stenosis, and this leads to the clinicians maintain the anatomical stenosis as their primary base for diagnosing and treatment of CAD. But studies have proven that the correlation between coronary artery stenosis (CAS) and myocardial ischemia was poor, the patient with CAS may be won’t suffer myocardial ischemia. Other studies have shown that only 50% patients get obviously CAS when they have myocardial microcirculation perfusion abnormalities. Hence, we need to master the myocardial microcirculation situation (functional significance of coronary artery) besides accurately diagnose the coronary artery anatomy situation (morphology significance of coronary artery). The invasive coronary angiography (ICA) is the same with CTCA who provides anatomical information and lacks of functional information, and as an invasive imaging method, ICA is gradually not accepted by patients. As a functional imaging method, myocardial perfusion imaging (MPI) especially the stress CT myocardial perfusion imaging receives increasing interest in coronary artery disease in these years.The commonly used MPI methods include single-photon emission computerized tomography (SPECT) MPI, stress echocardiography (SE), magnetic resonance MPI (MR-MPI) and CT-MPI. The SPECT-MPI has been the most commonly used non-invasive functional imaging modality for assessing of myocardial ischemia. By the statistics, there are about 1 billion patients after rest or stress SPECT-MPI examination every year in the world, therefore, SPECT-MPI is regarded as the gold standard in the diagnosis of CAD. Though the SPECT-MPI has been widely applied to the clinical, but its low spatial resolution, not able to show coronary artery anatomical details and insensitive to early diagnosis of CAD; the half-time of radioactive nuclide in the body is long which would endanger the human body health. SE refers to the use of exercising, pace-making and drug to make extra load on the heart, then testing the response from cardiovascular system by the ultrasonic, the patients who may suffer myocardial infarction will not allowed to use the above methods to increase heart load. But the low resolution, subjective factors and low repeatability by the ultrasound imaging limit the widely used in clinical. MR-MPI has good tissue contrast, high spatial and temporal resolution; it can present the size and extent of myocardial ischemia and also can reflect the myocardial microcirculation perfusion situation. But MR’s long time examination, expensive, difficult technology, not able to absolute quantitative analysis of myocardial blood flow and more contraindications limit its development. These three kinds of MPI methods (SPECT, SE and MR-MPI) have some shortages. CT-MPI owns high spatial and temporal resolution and has already been touted as the’one-stop’shop (OSS) to acquire functional and anatomical information of coronary artery stenosis accurately; the OSS method will be more and more important in the future. The first study of CT-MPI used electron beam CT (EBCT), EBCT has good temporal resolution and can obtain the CT-MP images from volume of interested; because of the limited spatial resolution, it unclear to show the details of CT-MP images, so it was eliminated. Multi-detector CT (MDCT) can cover the whole heart for CT-MPI, but the XY plane resolution is not high enough, and the additional edges of the detectors increased the MPI radiation dose. Therefore, EBCT and MDCT are not worth clinical promotion. In recent years, with the rapid development of CT technology, the world’s major CT producers have launched their own new products; especially when these high-end CT, such as 320-detector CT (320-DCT), Spectral CT, dual-source CT (DSCT) and Discorvery CT750 came out, more and more international and domestic academics used the high-end CT for research about.According to CT-MPI scan protocol, following an intravenous injection of iodinated contrast agent, continuous enhanced scan of the selected region of interest (ROI) myocardium slices of concern in a cine mode to acquire time density curve (TDC) of every pixels in interested slices, then the different mathematical models are used to calculate myocardial perfusion hemodynamic parameter map (MPHM), such as myocardial blood flow (MBF), myocardial blood volume (MBV) and mean transmit time (MTT); by analyzing these MPHMs, we can accurately diagnosis whether there is myocardial ischemia and the degree of myocardial ischemia. The commonly used mathematical models include non-deconvolution and deconvolution algorithms. The non-deconvolution includes the ’ignore the outflow vein method’ (also known as the moment method) and venous drainage method (also known as the maximum slope method); according to the induction-dilution theory (also known as Fick theory), the TDCs can be acquired through the contrast agent first pass through the volume of interest which used to calculate the MPHMs. The advantages of non-deconvolution methods are simple, easy to calculate and understand. The shortcomings are to ignore the venous outflow and no extravasation of contrast, so you need to fast and large dose injection of contrast, so that increases the risk of patients and lead to inconvenience of clinical application. Based on Fick theory and the hemodynamic of two-compartment model, the two known quantities contrast agent concentration (C(t)) and arterial input function (AIF) are deconvoluted to solve flow scaled residue function (K(t)), then using K(t) to compute the MPHMs. However, due to the sequential scan protocol, CT-MPI carries a larger radiation burden in comparison to that in conventional CT scans. A large number of studies have shown that excessive x-ray radiation exposure can induce leukemia, cancer and other genetic disorders. Now, the clinician use low dose CT scan protocol, such as lowering the milliampere-seconds (mAs), kVp and reducing the number of projections per rotation around the body in data acquisition may be a simple and cost-effective way. However, without adequate measures applied in image reconstruction, the associative CT-MP images quality would be degraded and cause of misdiagnosis due to the introduced serious photon noise. In guarantee under the premise of high-quality CT-MP images and not affects the doctor’s diagnosing, how to reduce X-ray radiation dose is an important and interesting research topic in the CT-MPI field.To date, various investigations about how to optimize CT-MP scanning process for dose reduction has been explored, all of them are divided into two categories. The first category methods include various hardware optimizations, image processing and advanced image reconstruction methods to obtain high quality CT-MP images. Such as optimizing the scan protocol and using of electrocardio-gating (ECG); Stenner et al. applied the partial scan artifact reduction (PSAR) technique to CT-MPI, the PSAR not only reducing the X-ray radiation dose significantly but also acquiring high temporal resolution CT-MP images. Without consideration of object motion from one time point to the next during CT-MPI, Speidel et al. proposed a highly constrained backprojection (HYPR) image processing method to directly reduce the low-dose MPCT image noise. Taking the motion artifacts into account due to heart motion in a full scan, J.C. Ramirez-Giraldo et al. investigated the partial scan reconstruction in CT-MPI using dynamic multi-detector computed tomography, and proposed a targeted spatial frequency filtration strategy (TSFF) to decrease partial scan reconstruction artifacts. Recently, Tao et al introduced statistical iterative reconstruction (SIR) framework to CT-MP reconstruction, this CT iterative reconstruction method which directly acts on sinogram data with remarkable improvements of noise reduction and streak artifacts suppression. The first category methods can yield reasonable CT-MP images, but they use unstable deconvolution algorithm to estimate MPHMs with acquired CT-MP images. Frindel and Fang research show that the robust deconvolution model must be used if you want to achieve diagnostician MPHMs. Hence, these robust deconvolution methods are called the second category methods which directly improve the estimation precision for MPHMs. Frindel et al. built a spatial and temporal regularization deconvolution model and applied it to perfusion-weighted magnetic resonance imaging (PW-MRI); Fang et al. presented a tensor total variation (TTV) regularization for low-dose cerebral perfusion CT (CPCT) deconvolution with noticeable gains in cerebral perfusion hemodynamic maps (CPHM) estimation. Of course, if the first category method is combined with the second category method and applied to CT-MPI, this will be another topic in future research.In this paper, the first category method and the second category method are studied, respectively. To sum up, the main work of this paper is as follow:1) To develop a motion adaptive sparsity prior (MASP) based algorithm for CT-MPI. Based on the analysis of extensive experiments, the difference between two adjacent frames is intensity gradient sparsity after motion correction to the acquired CT-MP images. And this spatial and temporal structured sparsity can be used as the priori information for the CT-MP image reconstruction; we develop a penalized weighted least-squares (PWLS) scheme by incorporating a MASP model to improve the image quality of CT-MPI. For simplicity, we refer to the proposed scheme as’PWLS-MASP’. Subsequently, we propose a modified alternating optimization method to minimize the associative objective function. The contribution of our work is four-fold:1) our method introduced the motion correction step to CT SIR; 2) the PWLS-MASP models both the spatial and temporal structured sparsity in CT-MP sequence images; 3) using a global convergence optimization method to minimize the associative objective function; 4) to validate and evaluate the effectiveness of the present PWLS-MASP algorithm, a modified XCAT phantom and preclinical porcine CT-MPI datasets were used, qualitative and quantitative evaluations demonstrated the proposed PWLS-MASP method can achieve promising gains in terms of noise reduction, streak artifacts mitigation, time density curves, and hemodynamic parameters.2) To develop an adaptive-weighted tensor total variation (AwTTV) regularization algorithm for CT-MPI. The AwTTV regularization algorithm regularizes the consistency of the solution by fusing the spatial correlation of the vascular structure and the temporal continuation of the blood flow signal. A modified alternating method was used in the proposed algorithm to minimize the associative objective function. Comperisons between the proposed method and other models (standard truncated singular value decomposition, sSVD; block-circulant truncated SVD, bSVD; Tikhonov regularization), the experimental results on the simulation XCAT and pig myocardial perfusion datasets demonstrated that the proposed method can significantly improve the quality of low-dose hemodynamic parameter maps by suppressing the noise and removing the streak-artifacts, and preserving the diagnosis value details in the myocardium.