Gaussian process regression for ultrasound scanline interpolation.

Purpose: In ultrasound imaging, interpolation is a key step in converting scanline data to brightness-mode (B-mode) images. Conventional methods, such as bilinear interpolation, do not fully capture the spatial dependence between data points, which leads to deviations from the underlying probability distribution at the interpolation points. Approach: We propose Gaussian process ( GP ) regression as an improved method for ultrasound scanline interpolation. Using ultrasound scanlines acquired from two different ultrasound scanners during in vivo trials, we compare the scanline conversion accuracy of three standard interpolation methods with that of GP regression, measuring the peak signal-to-noise ratio (PSNR) and mean absolute error (MAE) for each method. Results: The PSNR and MAE scores show that GP regression leads to more accurate scanline conversion compared to the nearest neighbor, bilinear, and cubic spline interpolation methods, for both datasets. Furthermore, limiting the interpolation window size of GP regression to 15 reduces computation time with minimal to no reduction in PSNR. Conclusions: GP regression quantitatively leads to more accurate scanline conversion and provides uncertainty estimates at each of the interpolation points. Our windowing method reduces the computational cost of using GP regression for scanline conversion.

Automatic extraction of the mitral valve chordae geometry for biomechanical simulation.

PURPOSE: Mitral valve computational models are widely studied in the literature. They can be used for preoperative planning or anatomical understanding. Manual extraction of the valve geometry on medical images is tedious and requires special training, while automatic segmentation is still an open problem. METHODS: We propose here a fully automatic pipeline to extract the valve chordae architecture compatible with a computational model. First, an initial segmentation is obtained by sub-mesh topology analysis and RANSAC-like model-fitting procedure. Then, the chordal structure is optimized with respect to objective functions based on mechanical, anatomical, and image-based considerations. RESULTS: The approach has been validated on 5 micro-CT scans with a graph-based metric and has shown an [Formula: see text] accuracy rate. The method has also been tested within a structural simulation of the mitral valve closed state. CONCLUSION: Our results show that the chordae architecture resulting from our algorithm can give results similar to experienced users while providing an equivalent biomechanical simulation.

Temporal enhancement of 2D color Doppler echocardiography sequences by fragment-based frame reordering and refinement.

PURPOSE: The goal of this study was to develop an algorithm that enhances the temporal resolution of two-dimensional color Doppler echocardiography (2D CDE) by reordering all the acquired frames and filtering out the frames corrupted by out-of-plane motion and arrhythmia. METHODS: The algorithm splits original frame sequence into the fragments based on the correlation with a reference frame. Then, the fragments are aligned temporally and merged into a resulting sequence that has higher temporal resolution. We evaluated the algorithm with 10 animal epicardial 2D CDE datasets of the right ventricle and compared it with the existing approaches in terms of resulting frame rate, image stability and execution time. RESULTS: We identified the optimal combination of alternatives for each step, which resulted in an increase in frame rate from 14 ± 0.87 to 238 ± 93 Hz. The average execution time was 7.23 ± 0.48 s in comparison with 0.009 ± 0.001 s for ECG gating and 1167.37 ± 587.85 s for flow reordering. Our approach demonstrated a significant (p < 0.01) increase in image stability compared with ECG gating and flow reordering. CONCLUSION: This work presents an offline algorithm for temporal enhancement of 2D CDE. Unlike previous frame reordering approaches, it can filter out-of-plane or corrupted frames, increasing the quality of the results, which substantially increases diagnostic value of 2D CDE. It can be used for high-frame-rate intraoperative imaging of intraventricular and valve regurgitant flows and is potentially modifiable for real-time use on ultrasound machines.

An intraoperative test device for aortic valve repair.

OBJECTIVE: Aortic valve repair is currently in transition from surgical improvisation to a reproducible operation and an option for many patients with aortic regurgitation. Our research efforts at improving reproducibility include development of methods for intraoperatively testing and visualizing the valve in its diastolic state. METHODS: We developed a device that can be intraoperatively secured in the transected aorta allowing the aortic root to be pressurized and the closed valve to be inspected endoscopically. Our device includes a chamber that can be pressurized with crystalloid solution and ports for introduction of an endoscope and measuring gauges. We show use of the device in explanted porcine hearts to visualize the aortic valve and to measure leaflet coaptation height in normal valves and in valves that have undergone valve repair procedures. RESULTS: The procedure of introducing and securing the device in the aorta, pressurizing the valve, and endoscopically visualizing the closed valve is done in less than 1 minute. The device easily and reversibly attaches to the aortic root and allows direct inspection of the aortic valve under conditions that mimic diastole. It enables the surgeon to intraoperatively study the valve immediately before repair to determine mechanisms of incompetence and immediately after the repair to assess competence. We also show its use in measuring valve leaflet coaptation height in the diastolic state. CONCLUSIONS: This device enables more relevant prerepair valve assessment and also enables a test of postrepair valve competence under physiological pressures.

High dynamic range ultrasound imaging.

PURPOSE: High dynamic range (HDR) imaging is a popular computational photography technique that has found its way into every modern smartphone and camera. In HDR imaging, images acquired at different exposures are combined to increase the luminance range of the final image, thereby extending the limited dynamic range of the camera. Ultrasound imaging suffers from limited dynamic range as well; at higher power levels, the hyperechogenic tissue is overexposed, whereas at lower power levels, hypoechogenic tissue details are not visible. In this work, we apply HDR techniques to ultrasound imaging, where we combine ultrasound images acquired at different power levels to improve the level of detail visible in the final image. METHODS: Ultrasound images of ex vivo and in vivo tissue are acquired at different acoustic power levels and then combined to generate HDR ultrasound (HDR-US) images. The performance of five tone mapping operators is quantitatively evaluated using a similarity metric to determine the most suitable mapping for HDR-US imaging. RESULTS: The ex vivo and in vivo results demonstrated that HDR-US imaging enables visualizing both hyper- and hypoechogenic tissue at once in a single image. The Durand tone mapping operator preserved the most amount of detail across the dynamic range. CONCLUSIONS: Our results strongly suggest that HDR-US imaging can improve the utility of ultrasound in image-based diagnosis and procedure guidance.

Fast image-based mitral valve simulation from individualized geometry.

BACKGROUND: Common surgical procedures on the mitral valve of the heart include modifications to the chordae tendineae. Such interventions are used when there is extensive leaflet prolapse caused by chordae rupture or elongation. Understanding the role of individual chordae tendineae before operating could be helpful to predict whether the mitral valve will be competent at peak systole. Biomechanical modelling and simulation can achieve this goal. METHODS: We present a method to semi-automatically build a computational model of a mitral valve from micro CT (computed tomography) scans: after manually picking chordae fiducial points, the leaflets are segmented and the boundary conditions as well as the loading conditions are automatically defined. Fast finite element method (FEM) simulation is carried out using Simulation Open Framework Architecture (SOFA) to reproduce leaflet closure at peak systole. We develop three metrics to evaluate simulation results: (i) point-to-surface error with the ground truth reference extracted from the CT image, (ii) coaptation surface area of the leaflets and (iii) an indication of whether the simulated closed leaflets leak. RESULTS: We validate our method on three explanted porcine hearts and show that our model predicts the closed valve surface with point-to-surface error of approximately 1 mm, a reasonable coaptation surface area, and absence of any leak at peak systole (maximum closed pressure). We also evaluate the sensitivity of our model to changes in various parameters (tissue elasticity, mesh accuracy, and the transformation matrix used for CT scan registration). We also measure the influence of the positions of the chordae tendineae on simulation results and show that marginal chordae have a greater influence on the final shape than intermediate chordae. CONCLUSIONS: The mitral valve simulation can help the surgeon understand valve behaviour and anticipate the outcome of a procedure.

Left Atrial Volumes and Strain in Healthy Children Measured by Three-Dimensional Echocardiography: Normal Values and Maturational Changes.

BACKGROUND: Assessment of left atrial (LA) size and function is important in a number of pediatric cardiac conditions including those affecting the diastolic performance of the left ventricle. Measurements of LA volume and strain by two-dimensional echocardiography rely upon inaccurate geometric assumptions and are hampered by out-of-plane motion. The objective of this study was to characterize LA volumes and strain by three-dimensional echocardiography in healthy children. METHODS: LA volumes and strain were retrospectively measured by three-dimensional echocardiography in healthy children with no known structural or functional heart disease using a commercial speckle-tracking package applied to the LA to compute maximum volume (Vmax), minimum volume (Vmin), ejection volume (LAEV), ejection fraction (LAEF), and the following components of global strain: 3D principal (3DS), longitudinal (GLS), and circumferential (GCS). RESULTS: The study population included 196 normal subjects (median age, 12 years; range, 4 days to 20.9 years). Vmax, Vmin, and LAEV increased with age and body surface area. Significant age-related declines were present in all measured functional indices including LAEF, 3DS, GLS, and GCS. Analysis of a subset of 50 subjects showed excellent agreement between Vmax derived by three-dimensional and two-dimensional biplane area-length method. Regression equations with standard deviations were generated to enable calculation of Z scores. CONCLUSIONS: LA volume and functional indices can be reliably calculated using a commercial three-dimensional analysis software. All components of LA strain decline modestly with age. Normative data generated in this study have the potential to greatly enhance the utility of three-dimensional echocardiography-derived measurements in a wide range of cardiac pathologies.

Automated detection of coarctation of aorta in neonates from two-dimensional echocardiograms.

Coarctation of aorta (CoA) is a critical congenital heart defect (CCHD) that requires accurate and immediate diagnosis and treatment. Current newborn screening methods to detect CoA lack both in sensitivity and specificity, and when suspected in a newborn, it must be confirmed using specialized imaging and expert diagnosis, both of which are usually unavailable at tertiary birthing centers. We explore the feasibility of applying machine learning methods to reliably determine the presence of this difficult-to-diagnose cardiac abnormality from ultrasound image data. We propose a framework that uses deep learning-based machine learning methods for fully automated detection of CoA from two-dimensional ultrasound clinical data acquired in the parasternal long axis view, the apical four chamber view, and the suprasternal notch view. On a validation set consisting of 26 CoA and 64 normal patients our algorithm achieved a total error rate of 12.9% (11.5% false-negative error and 13.6% false-positive error) when combining decisions of classifiers over three standard echocardiographic view planes. This compares favorably with published results that combine clinical assessments with pulse oximetry to detect CoA (71% sensitivity).

A Growth-Accommodating Implant for Paediatric Applications.

Medical implants of fixed size cannot accommodate normal tissue growth in children, and often require eventual replacement or in some cases removal, leading to repeated interventions, increased complication rates and worse outcomes. Implants that can correct anatomic deformities and accommodate tissue growth remain an unmet need. Here, we report the design and use of a growth-accommodating device for paediatric applications that consists of a biodegradable core and a tubular braided sleeve, with inversely related sleeve length and diameter. The biodegradable core constrains the diameter of the sleeve, and gradual core degradation following implantation enables sleeve and overall device elongation in order to accommodate tissue growth. By using mathematical modeling and ex vivo experiments using harvested swine hearts, we demonstrate the predictability and tunability of the behavior of the device for disease- and patient-specific needs. We also used the rat tibia and the piglet heart valve as two models of tissue growth to demonstrate that polymer degradation enables device expansion and growth accommodation in vivo.

Temporal enhancement of 3D echocardiography by frame reordering.

We describe a method to increase the frame rate for 3-dimensional ultrasound sequences of periodically moving cardiac structures by reordering the acquired volume series. The frame rate is especially important in studying intracardiac structures such as valve leaflet motion in which valve closing times are on the order of milliseconds. Current commercially available systems for volumetric ultrasound imaging are limited to approximately 10 to 20 volumes per second. Although this frame rate is sufficient for real-time observation of basic cardiac morphology, understanding cardiac dynamics requires faster frame rates. The presented work achieves higher frame rates by sampling over several beats and using a simultaneous electrocardiography signal to accurately place the frame within the cardiac cycle. The proposed method relies on periodicity of the heart motion and that within the temporal regions of highest velocity, the structural motions of interest have the lowest beat-to-beat variability.

Changes in left atrioventricular valve geometry after surgical repair of complete atrioventricular canal.

OBJECTIVE: The most common reason for late surgical reintervention after repair of complete atrioventricular canal defects is the development of left atrioventricular valve regurgitation. We sought to determine the changes in left atrioventricular valve geometry after surgical repair that may predispose to regurgitation. METHODS: Atrioventricular valve measurements were obtained by 2-dimensional echocardiography at 3 different time points (preoperative, early postoperative, and midterm postoperative [6-12 months]). Left atrioventricular valve annulus area and left ventricular volume were calculated; vena contracta of the regurgitant jet orifice was measured. All measurements were normalized relative to an appropriate power of body surface area. RESULTS: From January 2000 to January 2008, 101 patients with complete atrioventricular canal repair were included. Left atrioventricular valve annulus was noted to remodel from an elliptical shape to a circular shape after surgery. Left atrioventricular valve annulus area increased early postoperatively (systole: 4.1 ± 0.2 cm(2)/m(2) vs 6.1 ± 0.3 cm(2)/m(2), P < .001; diastole: 7.2 ± 0.4 cm(2)/m(2) vs 10.0 ± 0.5 cm(2)/m(2), P < .001, pre- vs postoperative, respectively). This increase was sustained in the midterm postoperative period (systole: 6.1 ± 0.3 cm(2)/m(2), P = .85, vs diastole: 10.0 ± 0.4 cm(2)/m(2), P = .78, early vs midterm postoperative). Left ventricular volume increased in the early and midterm postoperative periods compared with preoperative (systole: 16.9 ± 1.2 mL/m(2) vs 26.2 ± 1.7 mL/m(2), P < .001; diastole: 35.0 ± 2.4 mL/m(2) vs 52.5 ± 3.2 mL/m(2), P < .001). CONCLUSIONS: Complete atrioventricular canal repair leads to left atrioventricular valve annular shape change with increased area and circular shape. The change in left atrioventricular valve annulus shape appeared to be mainly due to increased circumference in the posterior free wall of the annulus. These findings may provide a mechanism for the progression of central regurgitation seen after complete atrioventricular canal repair and a potential solution.

Real-time image-based rigid registration of three-dimensional ultrasound.

Registration of three-dimensional ultrasound (3DUS) volumes is necessary in several applications, such as when stitching volumes to expand the field of view or when stabilizing a temporal sequence of volumes to cancel out motion of the probe or anatomy. Current systems that register 3DUS volumes either use external tracking systems (electromagnetic or optical), which add expense and impose limitations on acquisitions, or are image-based methods that operate offline and are incapable of providing immediate feedback to clinicians. This paper presents a real-time image-based algorithm for rigid registration of 3DUS volumes designed for acquisitions in which small probe displacements occur between frames. Described is a method for feature detection and descriptor formation that takes into account the characteristics of 3DUS imaging. Volumes are registered by determining a correspondence between these features. A global set of features is maintained and integrated into the registration, which limits the accumulation of registration error. The system operates in real-time (i.e. volumes are registered as fast or faster than they are acquired) by using an accelerated framework on a graphics processing unit. The algorithm's parameter selection and performance is analyzed and validated in studies which use both water tank and clinical images. The resulting registration accuracy is comparable to similar feature-based registration methods, but in contrast to these methods, can register 3DUS volumes in real-time.

Mitral annulus segmentation from four-dimensional ultrasound using a valve state predictor and constrained optical flow.

Measurement of the shape and motion of the mitral valve annulus has proven useful in a number of applications, including pathology diagnosis and mitral valve modeling. Current methods to delineate the annulus from four-dimensional (4D) ultrasound, however, either require extensive overhead or user-interaction, become inaccurate as they accumulate tracking error, or they do not account for annular shape or motion. This paper presents a new 4D annulus segmentation method to account for these deficiencies. The method builds on a previously published three-dimensional (3D) annulus segmentation algorithm that accurately and robustly segments the mitral annulus in a frame with a closed valve. In the 4D method, a valve state predictor determines when the valve is closed. Subsequently, the 3D annulus segmentation algorithm finds the annulus in those frames. For frames with an open valve, a constrained optical flow algorithm is used to the track the annulus. The only inputs to the algorithm are the selection of one frame with a closed valve and one user-specified point near the valve, neither of which needs to be precise. The accuracy of the tracking method is shown by comparing the tracking results to manual segmentations made by a group of experts, where an average RMS difference of 1.67±0.63mm was found across 30 tracked frames.

Patient-specific mitral leaflet segmentation from 4D ultrasound.

Segmenting the mitral valve during closure and throughout a cardiac cycle from four dimensional ultrasound (4DUS) is important for creation and validation of mechanical models and for improved visualization and understanding of mitral valve behavior. Current methods of segmenting the valve from 4DUS either require extensive user interaction and initialization, do not maintain the valve geometry across a cardiac cycle, or are incapable of producing a detailed coaptation line and surface. We present a method of segmenting the mitral valve annulus and leaflets from 4DUS such that a detailed, patient-specific annulus and leaflets are tracked throughout mitral valve closure, resulting in a detailed coaptation region. The method requires only the selection of two frames from a sequence indicating the start and end of valve closure and a single point near a closed valve. The annulus and leaflets are first found through direct segmentation in the appropriate frames and then by tracking the known geometry to the remaining frames. We compared the automatically segmented meshes to expert manual tracings for both a normal and diseased mitral valve, and found an average difference of 0.59 +/- 0.49 mm, which is on the order of the spatial resolution of the ultrasound volumes (0.5-1.0 mm/voxel).

Mitral annulus segmentation from 3D ultrasound using graph cuts.

The shape of the mitral valve annulus is used in diagnostic and modeling applications, yet methods to accurately and reproducibly delineate the annulus are limited. This paper presents a mitral annulus segmentation algorithm designed for closed mitral valves which locates the annulus in three-dimensional ultrasound using only a single user-specified point near the center of the valve. The algorithm first constructs a surface at the location of the thin leaflets, and then locates the annulus by finding where the thin leaflet tissue meets the thicker heart wall. The algorithm iterates until convergence metrics are satisfied, resulting in an operator-independent mitral annulus segmentation. The accuracy of the algorithm was assessed from both a diagnostic and surgical standpoint by comparing the algorithm's results to delineations made by a group of experts on clinical ultrasound images of the mitral valve, and to delineations made by an expert with a surgical view of the mitral annulus on excised porcine hearts using an electromagnetically tracked pointer. In the former study, the algorithm was statistically indistinguishable from the best performing expert (p=0.85) and had an average RMS difference of 1.81+/-0.78 mm to the expert average. In the latter, the average RMS difference between the algorithm's annulus and the electromagnetically tracked points across six hearts was 1.19+/-0.17 mm .

Beating-heart mitral valve suture annuloplasty under real-time three-dimensional echocardiography guidance: an ex vivo study.

We are developing an alternative mitral valve suture annuloplasty technique on the beating-heart under real-time three-dimensional echocardiography (RT3DE) guidance. The purpose of this initial study was to evaluate a feasibility of this technique using commercially available suturing devices (Sutur Tek Endo 360-degree, Sutur Tek Inc, North Chelmsford, MA, USA). Isolated porcine hearts (n=10) were mounted in a water-filled tank and attached to an ex vivo pulse simulation device, where varying left ventricle pressures with associated valve motion were generated by pulsatile flow through an apical cannula. The suturing device was inserted through the left atrium. Intra-annular (De Vega type) suture annuloplasty was performed under RT3DE guidance. The procedure was successfully performed in all cases. The diameter of the annulus was effectively reduced (85.5+/-4.2% of original antero-posterior dimension, 86.7+/-6.1% of original transverse dimension). The number of tissue bites was 7.4+/-0.8. The maximum distance between the annulus and sutures placed was 1.1 mm. The total procedure time was 9.4+/-2.4 min. There was no collateral tissue injury in any of the cases. This ex vivo study demonstrates the feasibility of beating-heart mitral valve suture annuloplasty under RT3DE guidance.

Force Tracking with Feed-Forward Motion Estimation for Beating Heart Surgery.

The manipulation of fast moving, delicate tissues in beating heart procedures presents a considerable challenge to the surgeon. A robotic force tracking system can assist the surgeon by applying precise contact forces to the beating heart during surgical manipulation. Standard force control approaches cannot safely attain the required bandwidth for this application due to vibratory modes within the robot structure. These vibrations are a limitation even for single degree of freedom systems driving long surgical instruments. These bandwidth limitations can be overcome by incorporating feed-forward motion terms in the control law. For intracardiac procedures, the required motion estimates can be derived from 3D ultrasound imaging. Dynamic analysis shows that a force controller with feed-forward motion terms can provide safe and accurate force tracking for contact with structures within the beating heart. In vivo validation confirms that this approach confers a 50% reduction in force fluctuations when compared to a standard force controller and a 75% reduction in fluctuations when compared to manual attempts to maintain the same force.

Mitral Annulus Segmentation from Three-Dimensional Ultrasound.

An accurate and reproducible segmentation of the mitral valve annulus from 3D ultrasound is useful to clinicians and researchers in applications such as pathology diagnosis and mitral valve modeling. Current segmentation methods, however, are based on 2D information, resulting in inaccuracies and a lack of spatial coherence. We present a segmentation algorithm which, given a single user-specified point near the center of the valve, uses max-flow and active contour methods to delineate the annulus geometry in 3D. Preliminary comparisons to manual segmentations and a sensitivity study show the algorithm is both accurate and robust.