CT-based diagnosis of diffuse coronary artery disease on the basis of scaling power laws

The feasibility of early detecting coronary artery disease using deep learning-based algorithm based on electrocardiography

BACKGROUND

Coronary Artery Disease (CAD) is a major cause of morbidity and mortality, yet it is frequently asymptomatic in the early stages and hence goes undetected.

OBJECTIVE

We aimed to develop a novel artificial intelligence-based approach for early detection of CAD patients based solely on electrocardiogram (ECG).

METHODS

This study included patients with suspected CAD who had standard 10-s resting 12-lead ECGs and coronary computed tomography angiography (cCTA) results within 4 weeks or less. The ECG and cCTA data from the same patient were matched based on their hospitalization or outpatient ID. All matched data pairs were then randomly divided into training, validation dataset for model development based on convolutional neural network (CNN) and test dataset for model evaluation. The accuracy (Acc), specificity (Spec), sensitivity (Sen), positive predictive value (PPV), negative predictive value (NPV) and area under the receiver operating characteristic curve (AUC) of the model were calculated by using the test dataset.

RESULTS

In the test dataset, the model for detecting CAD achieved an AUC of 0.75 (95% CI, 0.73 to 0.78) with an accuracy of 70.0%. Using the optimal cut-off point, the CAD detection model had sensitivity of 68.7%, specificity of 70.9%, positive predictive value (PPV) of 61.2%, and negative predictive value (NPV) of 77.2%. Our study demonstrates that a well-trained CNN model based solely on ECG could be considered an efficient, low-cost, and noninvasive method of assisting in CAD detection.

One-stop patient-specific myocardial blood flow quantification technique based on allometric scaling law

Establishing a patient-specific and non-invasive technique to derive blood flow as well as coronary structural information from one single cardiac CT imaging modality. 336 patients with chest pain or ST segment depression on electrocardiogram were retrospectively enrolled. All patients underwent adenosine-stressed dynamic CT myocardial perfusion imaging (CT-MPI) and coronary computed tomography angiography (CCTA) in sequence. Relationship between myocardial mass (M) and blood flow (Q), defined as log(Q) = b · log(M) + log(Q₀), was explored based on the general allometric scaling law. We used 267 patients to obtain the regression results and found strong linear relationship between M (gram) and Q (mL/min) (b = 0.786, log(Q₀) = 0.546, r = 0.704; p < 0.001). We Also found this correlation was applicable for patients with either normal or abnormal myocardial perfusion (p < 0.001). Datasets from the other 69 patients were used to validate this M-Q correlation and found the patient-specific blood flow could be accurately estimated from CCTA compared to that measured from CT-MPI (146.480 ± 39.607 vs 137.967 ± 36.227, r = 0.816, and 146.480 ± 39.607 vs 137.967 ± 36.227, r = 0.817, for the left ventricle region and LAD-subtended region, respectively, all unit in mL/min). In conclusion, we established a technique to provide general and patient-specific myocardial mass-blood flow correlation obeyed to allometric scaling law. Blood flow information could be directly derived from structural information acquired from CCTA.

PERSONALIZED FLOW DIVISION METHOD BASED ON THE LEFT-RIGHT CORONARY CROSS-SECTIONAL AREA

This paper proposes a personalized method to estimate blood flow distribution based on the cross-sectional area of the left-right coronary artery openings. According to the cross-sectional area of the left-right coronary artery in 30 cases, a personalized flow distribution model was derived. A 0D/3D geometric multiscale model was used for the numerical simulation of FFR. To evaluate the accuracy of the cross-sectional area method, invasive FFR was used as the standard. The diagnostic efficiency of the proposed method was verified through the simulation results of the volume and the fixed ratio methods. The flow of the left-right coronary artery was proportional to the 3/4 power of the cross-sectional area. The 95% LOA between the cross-sectional area method, volume method, fixed ratio method and FFR were [Formula: see text]0.06 ([[Formula: see text]0.22, 0.10]), [Formula: see text]0.03 ([[Formula: see text]0.35, 0.28]), and [Formula: see text]0.05 ([[Formula: see text]0.30, 0.20]), the accuracy values were 94.44%, 77.78%, and 77.78%, respectively. Flow distribution based on the cross-sectional area represents the supply and demand relationship of the myocardium. The flow of the left-right coronary arteries is proportional to the 3/4 exponent of the cross-sectional area, which affects the accuracy of FFRCT by affecting the exit boundary conditions of the 0D/3D model.

Allometric scaling patterns among the human coronary artery tree, myocardial mass, and coronary artery flow

Human coronary artery tree is a physiological transport system for oxygen and vital materials through a hierarchical vascular network to match the energy demands of myocardium, which has the highest oxygen extraction ratio among body organs and heavily depends on the blood flow for its energy supply. Therefore, it would be reasonable to expect that the key design principle of this arterial network is to minimize energy expenditure, which can be described by allometric scaling law. We enrolled patients who underwent coronary computed tomography angiography without obstructive lesion. The cumulative arterial length (L), volume (V), and diameter (D) in relation to the artery-specific myocardial mass (M) were assessed. Flow rate (Q) was computed using quantitative flow ratio (QFR) measurement in patients who underwent invasive angiography. A total of 638 arteries from 43 patients (mean age 61 years, male gender 65%) were analyzed. A significant power-law relationship was found among L-M, V-M, D-M, V-L, D-L, and V-D, and also among Q-M, Q-L, Q-V, and Q-D in 106 arteries interrogated with QFR (p < .001, all). Our results suggest that the fundamental design principle of the human coronary arterial network may follow allometric scaling law.

Vertebral Artery Stenoses Contribute to the Development of Diffuse Plaques in the Basilar Artery

Vertebral artery (VA) stenosis is relevant to a high early risk of recurrent stroke and basilar artery (BA) is the most common intracranial site of atherosclerotic lesions. It is important to show predictive risk factors for transient ischemic attack (TIA) or posterior infarctions. The aim of the study is to investigate morphometry and hemodynamics in intracranial vertebral and basilar arteries of health and diseased patients to enhance the risk assessment. Based on the geometrical model reconstructed from CTA images in 343 patients, a transient three-dimensional computational model was used to determine the hemodynamics. Patients were classified in symmetric, asymmetric, hypoplastic, and stenotic groups while patients in the stenotic group were divided into unilateral, bilateral, bifurcation, and tandem stenotic sub-groups. Patients in bilateral, bifurcation, and tandem stenotic sub-groups had significantly lower basilar artery diameters than other groups. Patients in the stenotic group had significantly higher surface area ratio (SAR) of high time-averaged wall shear stress gradient (TAWSSG) and higher incidence of TIAs or posterior infarctions than other groups while patients in the tandem stenotic sub-group had the highest values (SAR-TAWSSG of 57 ± 22% and TIAs or posterior infarction incidence of 54%). The high SAR-TAWSSG is predisposed to induce TIAs or posterior infarction.

Diagnostic performance of a fast non-invasive fractional flow reserve derived from coronary CT angiography: an initial validation study

AIM

To validate the computed tomography (CT)-derived fractional flow reserve (FFR_CT) that was computed by new, fast, automatic software and to compare the diagnostic accuracy between FFR_CT and stenosis diagnosed at coronary CT angiography (CCTA).

MATERIALS AND METHODS

A total of 110 patients (76 males, 59±9 years) and 125 vessels underwent CCTA. FFR_CT was computed by fast automatic software and compared with invasive FFR. The diagnostic performance between FFR_CT and CCTA-diagnosed stenosis were compared on the per-patient and per-vessel level.

RESULTS

The computational time of FFR_CT is 10±5 minutes (averaged over 125 vessels). The FFR_CT has a good correlation with invasive FFR (r=0.59, p<0.0001) with a small bias of -0.02 (-0.26-0.23). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of FFR_CT were 76.5, 89.5, 89.7, and 76.1% on a vessel level. The area under the receiver operating characteristic curve of FFR_CT was higher than CCTA-diagnosed stenosis (0.82 versus 0.72, P=0.034).

CONCLUSION

The computation of FFR_CT is possible and reliable when using the new, fast, automatic software first employed in the present clinical study. The FFR_CT has a good correlation with invasive FFR and provides better diagnostic performance than CCTA-diagnosed stenosis.

Effect of the ratio of coronary arterial lumen volume to left ventricle myocardial mass derived from coronary CT angiography on fractional flow reserve

BACKGROUND

We hypothesize that in patients with suspected coronary artery disease (CAD), lower values of the ratio of total epicardial coronary arterial lumen volume to left ventricular myocardial mass (V/M) result in lower fractional flow reserve (FFR).

METHODS

V/M was computed in 238 patients from the NXT trial who underwent coronary computed tomography angiography (CTA), quantitative coronary angiography (QCA) and FFR measurement in 438 vessels. Nitroglycerin was administered prior to CT, QCA and FFR acquisition. The V/M ratio was quantified on a patient-level from CT image data by segmenting the epicardial coronary arterial lumen volume (V) and the left ventricular myocardial mass (M). Calcified and noncalcified plaque volumes were quantified using semi-automated software.

RESULTS

The median value of V/M (18.57 mm³/g) was used to define equal groups of low and high V/M patients. Patients with low V/M had greater diameter stenosis by QCA, more plaque and lower FFR (0.80 ± 0.12 vs. 0.87 ± 0.08; P < 0.0001) than those with high V/M. A total of 365 vessels in 202 patients had QCA stenosis ≤50% and measured FFR. In these patients, those with low V/M had higher percent diameter stenosis by QCA, greater total plaque volume and lower FFR (0.81 ± 0.12 vs. 0.88 ± 0.07; P < 0.0001) than those with high V/M. In multivariate logistic regression analysis, V/M was an independent predictor of FFR ≤0.80 (all p-values < 0.001).

CONCLUSIONS

Patients with a low V/M ratio have lower FFR overall and in non-obstructive CAD, independent of plaque measures.

Intraspecific scaling laws are preserved in ventricular hypertrophy but not in heart failure

It is scientifically and clinically important to understand the structure-function scaling of coronary arterial trees in compensatory (e.g., left and right ventricular hypertrophy, LVH and RVH) and decompensatory vascular remodeling (e.g., congestive heart failure, CHF). This study hypothesizes that intraspecific scaling power laws of vascular trees are preserved in hypertrophic hearts but not in CHF swine hearts. To test the hypothesis, we carried out the scaling analysis based on morphometry and hemodynamics of coronary arterial trees in moderate LVH, severe RVH, and CHF compared with age-matched respective control hearts. The scaling exponents of volume-diameter, length-volume, and flow-diameter power laws in control hearts were consistent with the theoretical predictions (i.e., 3, 7/9, and 7/3, respectively), which remained unchanged in LVH and RVH hearts. The scaling exponents were also preserved with an increase of body weight during normal growth of control animals. In contrast, CHF increased the exponents of volume-diameter and flow-diameter scaling laws to 4.25 ± 1.50 and 3.15 ± 1.49, respectively, in the epicardial arterial trees. This study validates the predictive utility of the scaling laws to diagnose vascular structure and function in CHF hearts to identify the borderline between compensatory and decompensatory remodeling.

Hemodynamics in Coronary Arterial Tree of Serial Stenoses

Serial segmental narrowing frequently occurs in humans, which alters coronary hemodynamics and further affects atherosclerotic progression and plaque formation. The objective of this study was to understand the distribution of hemodynamic parameters in the epicardial left main coronary arterial (LMCA) tree with serial stenoses reconstructed from patient computer tomography angiography (CTA) images. A finite volume method was used in conjunction with the inlet pressure wave and outlet flow resistance. The time-averaged wall shear stress (TAWSS) and oscillatory shear index (OSI) were determined from the flow field. A stenosis at a mother vessel mainly deteriorated the hemodynamics near the bifurcation while a stenosis at a daughter vessel affected the remote downstream bifurcation. In comparison with a single stenosis, serial stenoses increased the peak pressure gradient along the main trunk of the epicardial left anterior descending arterial tree by > 50%. An increased distance between serial stenoses further increased the peak pressure gradient. These findings have important implications on the diagnosis and treatment of serial coronary stenoses.

Growth, ageing and scaling laws of coronary arterial trees

Despite the well-known design principles of vascular systems, it is unclear whether the vascular arterial tree obeys some scaling constraints during normal growth and ageing in a given species. Based on the micro-computed tomography measurements of coronary arterial trees in mice at different ages (one week to more than eight months), we show a constant exponent of 3/4, but age-dependent scaling coefficients in a length-volume scaling law (Lc=K(length-volume) · Vc³/⁴; Lc is the crown length, Vc is the crown volume, K(length-volume) is the age-dependent scaling coefficient) during normal growth and ageing. The constant 3/4 exponent represents the self-similar fractal-like branching pattern (i.e. basic mechanism to regulate the development of vascular trees within a species), whereas the age-dependent scaling coefficients characterize the structural growth or resorption of vascular trees during normal growth or ageing, respectively. This study enhances the understanding of age-associated changes in vascular structure and function.

Cardiac radiology: centenary review

During the past century, cardiac imaging technologies have revolutionized the diagnosis and treatment of acquired and congenital heart disease. Many important contributions to the field of cardiac imaging were initially reported in Radiology. The field developed from the early stages of cardiac imaging, including the use of coronary x-ray angiography and roentgen kymography, to nowadays the widely used echocardiographic, nuclear medicine, cardiac computed tomographic (CT), and magnetic resonance (MR) applications. It is surprising how many of these techniques were not recognized for their potential during their early inception. Some techniques were described in the literature but required many years to enter the clinical arena and presently continue to expand in terms of clinical application. The application of various CT and MR contrast agents for the diagnosis of myocardial ischemia is a case in point, as the utility of contrast agents continues to expand the noninvasive characterization of myocardium. The history of cardiac imaging has included a continuous process of advances in our understanding of the anatomy and physiology of the cardiovascular system, along with advances in imaging technology that continue to the present day.

Mild anastomotic stenosis in patient-specific CABG model may enhance graft patency: a new hypothesis

It is well known that flow patterns at the anastomosis of coronary artery bypass graft (CABG) are complex and may affect the long-term patency. Various attempts at optimal designs of anastomosis have not improved long-term patency. Here, we hypothesize that mild anastomotic stenosis (area stenosis of about 40–60%) may be adaptive to enhance the hemodynamic conditions, which may contribute to slower progression of atherosclerosis. We further hypothesize that proximal/distal sites to the stenosis have converse changes that may be a risk factor for the diffuse expansion of atherosclerosis from the site of stenosis. Twelve (12) patient-specific models with various stenotic degrees were extracted from computed tomography images using a validated segmentation software package. A 3-D finite element model was used to compute flow patterns including wall shear stress (WSS) and its spatial and temporal gradients (WSS gradient, WSSG, and oscillatory shear index, OSI). The flow simulations showed that mild anastomotic stenosis significantly increased WSS (>15 dynes⋅cm⁻²) and decreased OSI (<0.02) to result in a more uniform distribution of hemodynamic parameters inside anastomosis albeit proximal/distal sites to the stenosis have a decrease of WSS (<4 dynes⋅cm⁻²). These findings have significant implications for graft adaptation and long-term patency.