A learning-based, region of interest-tracking algorithm for catheter detection in echocardiography.

Echocardiography (echo) is gaining popularity to guide the catheter during surgical procedures. However, it is difficult to discern the catheter tip in echo even with an acoustically active catheter. An acoustically active catheter is detected for the first time in cardiac echo images using two methods. First, a convolutional neural network (CNN) model was trained to detect the region of interest (ROI), the interior of the left ventricle, containing the catheter tip. Color intensity difference detection technique was implemented on the ROI to detect the catheter. This method succeeded in detecting the catheter without any manual input on 94% and 57% of long- and short-axis projections, respectively. Second, several tracking methods were implemented and tested. Given the manually identified initial positions of the catheter, the tracking methods could distinguish between the target (catheter tip) and the surrounding on the rest of the frames. Combining the two techniques, for the first time, resulted in an automatic, robust, and fast method for catheter detection in echo images.

The Effects of Implantation Orientation of a Bileaflet Mechanical Heart Valve in an Anatomic Left Ventricle-Aorta Configuration.

We have performed three-dimensional high-resolution numerical simulations of a bi-leaflet mechanical heart valve implanted at different orientations in an anatomic left ventricle-aorta obtained from magnetic resonance imaging of a volunteer. The thoroughly validated overset curvilinear-immersed boundary fluid-structure interaction flow solver is used in which the aorta and left ventricle (LV) are discretized with boundary-conforming and nonconforming curvilinear grids, respectively. The motion of the left ventricle wall is prescribed based on a lumped parameter model while the motion of the leaflets is calculated using a strongly coupled fluid-structure interaction algorithm enhanced with Aitken convergence technique. We carried out simulations for three valve orientations, which differ from each other by 45 deg, and compared the leaflet motion and flow field for multiple cycles. Our results show reproducible and relatively symmetrical opening for all valve orientations. The presence of small-scale vortical structures after peak systole causes significant cycle-to-cycle variations in valve kinematics during the closing phase for all valve orientations. Furthermore, our results show that valve orientation does not have a significant effect on the distribution of viscous shear stress in the ascending aorta. Additionally, two different mathematical activation models including linear level of activation and Soares model are used to quantify the platelet activation in the ascending aorta. The results show that the valve orientation does not significantly affect (less than 8%) the total platelet activation in the ascending aorta.

Automatic segmentation of the left ventricle in echocardiographic images using convolutional neural networks.

BACKGROUND: Two-dimensional echocardiography (2D echo) is the most widely used non-invasive imaging modality due to its fast acquisition time, low cost, and high temporal resolution. Boundary identification of left ventricle (LV) in 2D echo, i.e., image segmentation, is the first step to calculate relevant clinical parameters. Currently, LV segmentation in 2D echo is primarily conducted semi-manually. A fully-automatic segmentation of the LV wall needs further development. METHODS: We evaluated the performance of the state-of-the-art convolutional neural networks (CNNs) for the segmentation of 2D echo images from 6 standard projections of the LV. We used two segmentation algorithms: U-net and segAN. The models were trained using an in-house dataset, which consists of 1,649 porcine images from 6 to 8 different pigs. In addition, a transfer learning approach was used for the segmentation of long-axis projections by training models with our database based on the previously trained weights obtained from Cardiac Acquisitions for Multi-structure Ultrasound Segmentation (CAMUS) dataset. The models were tested on a separate set of images from two other pigs by computing several metrics. The segmentation process was combined with a 3D reconstruction framework to quantify the physiological indices such as LV volumes and ejection fraction (EF). RESULTS: The average dice metric for the LV cavity was 0.90 and 0.91 for the U-net and segAN, respectively, which was higher than 0.82 for the level-set (P value: 3.31×10-25). The average Hausdorff distance for the LV cavity was 2.71 mm and 2.82 mm for the U-net and segAN, respectively, which was lower than 3.64 mm for the level-set (P value: 4.86×10-16). The LV shapes and volumes obtained using the CNN segmentation models were in good agreement with the results segmented by the experts. In addition, the differences of the calculated physiological parameters between two 3D reconstruction models segmented by the experts and CNNs were less than 15%. CONCLUSIONS: The results showed that both CNN models achieve higher performance on LV segmentation than the level-set method. The error of the reconstruction from automatic segmentation compared to the expert segmentation is less than 15%, which is within the 20% error of echo compared to the gold standard.

Fluid-structure coupled biotransport processes in aortic valve disease.

Biological transport processes near the aortic valve play a crucial role in calcific aortic valve disease initiation and bioprosthetic aortic valve thrombosis. Hemodynamics coupled with the dynamics of the leaflets regulate these transport patterns. Herein, two-way coupled fluid-structure interaction (FSI) simulations of a 2D bicuspid aortic valve and a 3D mechanical heart valve were performed and coupled with various convective mass transport models that represent some of the transport processes in calcification and thrombosis. Namely, five different continuum transport models were developed to study biochemicals that originate from the blood and the leaflets, as well as residence-time and flow stagnation. Low-density lipoprotein (LDL) and platelet activation were studied for their role in calcification and thrombosis, respectively. Coherent structures were identified using vorticity and Lagrangian coherent structures (LCS) for the 2D and 3D models, respectively. A very close connection between vortex structures and biochemical concentration patterns was shown where different vortices controlled the concentration patterns depending on the transport mechanism. Additionally, the relationship between leaflet concentration and wall shear stress was revealed. Our work shows that blood flow physics and coherent structures regulate the flow-mediated biological processes that are involved in aortic valve calcification and thrombosis, and therefore could be used in the design process to optimize heart valve replacement durability.

A hybrid echocardiography-CFD framework for ventricular flow simulations.

Image-based CFD is a powerful tool to study cardiovascular flows while 2D echocardiography (echo) is the most widely used noninvasive imaging modality for the diagnosis of heart disease. Here, echo is combined with CFD, that is, an echo-CFD framework, to study ventricular flows. To achieve this, the previous 3D reconstruction from multiple 2D echo at standard cross sections is extended by: (a) reconstructing aortic and mitral valves from 2D echo and closing the left-ventricle (LV) geometry by approximating a superior wall; (b) incorporating the physiological assumption of the fixed apex as a reference (fixed) point in the 3D reconstruction; and (c) incorporating several smoothing algorithms to remove the nonphysical oscillations (ringing) near the basal section. The method is applied to echo from a baseline LV and one after inducing acute myocardial ischemia (AMI). The 3D reconstruction is validated by comparing it against a reference reconstruction from many echo sections while flow simulations are validated against the Doppler ultrasound velocity measurements. The sensitivity study shows that the choice of the smoothing algorithm does not change the flow pattern inside the LV. However, the presence of the mitral valve can significantly change the flow pattern during the diastole phase. In addition, the abnormal shape of a LV with AMI can drastically change the flow during diastole. Furthermore, the hemodynamic energy loss, as an indicator of the LV pumping performance, for different test cases is calculated, which shows a larger energy loss for a LV with AMI compared to the baseline one.

Comparison of platelet activation through hinge vs bulk flow in bileaflet mechanical heart valves.

Bileaflet mechanical heart valves (BMHVs) are prone to thromboembolic complications which are believed to be initiated by platelet activation. Platelets are activated by non-physiologic shear stresses in the bulk flow or the leakage/hinge flow, whose contributions has yet to be quantified. Here, the contribution of bulk and hinge flows to the activation of platelets in BMHVs is quantified for the first time by performing simulations of the flow through a BMHV and resolving the hinge by overset grids (one grid for the bulk flow and two for the hinge regions coupled together using one-way and two-way interpolation). It was found that two-way coupling is essential to obtain correct hinge flow features. The platelet activation through the hinge for two gap sizes (250 and 150 µm) is compared to the activation in the bulk flow using two platelet activation models to ensure the consistency of the observed trends. The larger gap has a higher total activation, but a better washout ability due to higher velocities. The maximum shear stress observed in the bulk flow (∼320dyne/cm2) is much smaller than the hinge (∼1000dyne/cm2). However, the total activation by the bulk flow is found to be several folds higher than by the hinge/leakage flow. This is mainly due to the higher flow rate of the bulk flow which exposes much more platelets to shear stress than the leakage flow.

Platelet activation of mechanical versus bioprosthetic heart valves during systole.

Thrombus formation is a major concern for recipients of mechanical heart valves (MHVs), which requires them to take anticoagulant drugs for the rest of their lives. Bioprosthetic heart valves (BHVs) do not require life-long anticoagulant therapy but deteriorate after 10-15years. The thrombus formation is initiated by the platelet activation which is thought to be mainly generated in MHVs by the flow through the hinge and the leakage flow during the diastole. However, our results show that the activation in the bulk flow during the systole phase might play an essential role as well. This is based on our results obtained by comparing the thrombogenic performance of a MHV and a BHV (as control) in terms of shear induced platelet activation under exactly the same conditions. Three different mathematical activation models including linear level of activation, damage accumulation, and Soares model are tested to quantify the platelet activation during systole using the previous simulations of the flow through MHV and BHV in a straight aorta under the same physiologic flow conditions. Results indicate that the platelet activation in the MHV at the beginning of the systole phase is slightly less than the BHV. However, at the end of the systole phase the platelet activation by the bulk flow for the MHV is several folds (1.41, 5.12, and 2.81 for linear level of activation, damage accumulation, and Soares model, respectively) higher than the BHV for all tested platelet activation models.