Gabriel Kiss

Three-Dimensional Echocardiography in the Evaluation of Global and Regional Function in Patients with Recent Myocardial Infarction: A Comparison with Magnetic Resonance Imaging

We aimed to compare three‐dimensional (3D) and two‐dimensional (2D) echocardiography in the evaluation of patients with recent myocardial infarction (MI), using late‐enhancement magnetic resonance imaging (LE‐MRI) as a reference method. Echocardiography and LE‐MRI were performed approximately 1 month after first‐time MI in 58 patients. Echocardiography was also performed on 35 healthy controls. Left ventricular (LV) ejection fraction by 3D echocardiography (3D‐LVEF), 3D wall‐motion score (WMS), 2D‐WMS, 3D speckle tracking–based longitudinal, circumferential, transmural and area strain, and 2D speckle tracking–based longitudinal strain (LS) were measured. The global correlations to infarct size by LE‐MRI were significantly higher (P < 0.03) for 3D‐WMS and 2D‐WMS compared with 3D‐LVEF and the 4 different measurements of 3D strain, and 2D global longitudinal strain (GLS) was more closely correlated to LE‐MRI than 3D GLS (P < 0.03). The segmental correlations to infarct size by LE‐MRI were also significantly higher (P < 0.04) for 3D‐WMS, 2D‐WMS, and 2D LS compared with the other indices. Three‐dimensional WMS showed a sensitivity of 76% and a specificity of 72% for identification of LV infarct size >12%, and a sensitivity of 73% and a specificity of 95% for identification of segments with transmural infarct extension. Three‐dimensional WMS and 2D gray‐scale echocardiography showed the strongest correlations to LE‐MRI. The tested 3D strain method suffers from low temporal and spatial resolution in 3D acquisitions and added diagnostic value could not be proven.

Automatic coupled segmentation of endo- and epicardial borders in 3D echocardiography

In this paper, we present an extension of a computationally efficient Kalman filter based tracking framework to allow simultaneous tracking of several deformable models in volumetric data. The models are coupled through shared transform nodes in a hierarchical structure to enforce common position, orientation and scaling, but are allowed to alter shape independently. Automatic tracking of the endo- and epicardial border using two Doo-Sabin subdivision surfaces in 3D echocardiography serves as exemplary application. Fully automatic endo- and epicardial surface is demonstrated in a simulation and 5 in-vivo recordings of high image quality. The estimated myocardial volumes is overestimated by 4.2 ml (7.0%) compared to ground truth in the simulation, whereas the volumes in the in-vivo recordings are on average underestimated by 10.7 ml (8.0%) compared to independent reference segmentations.

Strain rate imaging combined with wall motion analysis gives incremental value in direct quantification of myocardial infarct size

BACKGROUND The study aimed to evaluate the diagnostic accuracy of a new method for direct echocardiographic quantification of the myocardial infarct size, using late enhancement magnetic resonance imaging (LE-MRI) as a reference method. METHODS AND RESULTS Echocardiography and LE-MRI were performed on average 31 days after first-time myocardial infarction in 58 patients. Echocardiography was also performed on 35 healthy controls. Direct echocardiographic quantification of the infarct size was based on automated selection and quantification of areas with hypokinesia and akinesia from colour-coded strain rate data, with manual correction based on visual wall motion analysis. The left ventricular (LV) ejection fraction, speckle-tracking-based longitudinal global strain, wall motion score index (WMSI), longitudinal systolic motion and velocity, and the ratio of early mitral inflow velocity to mitral annular early diastolic velocity were also measured by echocardiography. The area under the receiver-operating characteristic curves for the identification of the infarct size >12% by LE-MRI was 0.84, using the new method for direct echocardiographic quantification of the infarct size. The new method showed significantly a higher correlation with the infarct size by LE-MRI both at the global (r = 0.81) and segmental (r = 0.59) level compared with other indices of LV function. CONCLUSION Direct quantification of the percentage infarct size by strain rate imaging combined with wall motion analysis yields high diagnostic accuracy and better correlation to LE-MRI compared with other echocardiographic indices of global LV function. Echocardiography performed ~1 month after myocardial infarction showed ability to identify the patients with the infarct size >12%.

GPU volume rendering in 3D echocardiography: Real-time pre-processing and ray-casting

Since real-time acquisition of 3D echocardiographic data is achievable in practice, many volume rendering algorithms have been proposed for visualization purposes. However, due to the large amounts of data and computations involved a tradeoff between image quality and computational efficiency has to be made. The main goal of our study was to generate high quality volume renderings in real-time, by implementing preprocessing and ray-casting algorithms directly on the GPU. Furthermore the advantage of combining a-priori anatomic and functional information with the volume rendered image was also investigated. The proposed algorithms were implemented both in CUDA and OpenCL and validated on patient datasets acquired using a GE Vivid7 Dimensions system. Assuming a 512×512 pixels output resolution, average running times of 4.2 ms/frame are achievable on high-end graphics systems. Furthermore a good correspondence between wall thickening and segmental longitudinal strain values was visually observed. By implementing ray-casting on the GPU, the overall processing time is significantly reduced, thus making real-time interactive 3D volume rendering feasible. Combining anatomical and functional information allows for a quick visual assessment of a given case.

Combining edge detection with speckle-tracking for cardiac strain assessment in 3D echocardiography

In this paper, we extend a computationally efficient framework for tracking of deformable subdivision surfaces in 3D echocardiography with speckle-tracking measurements to track material points. Tracking is performed in a sequential state-estimation fashion, using an extended Kalman filter to update a subdivision surface in a two-step process: Edge-detection is first performed to update the model for changes in shape and position, followed by a second update based on displacement vectors from speckle-tracking with 3D block-matching. The latter speckle-tracking update will only have to correct for residual deformations after edge-detection. This both leads to increased accuracy and computational efficiency compared to usage of speckle-tracking alone. Automatic tracking is demonstrated in a 3D echocardiography simulation of an infarcted ventricle. The combination of edge-detection and speckle-tracking consistently improved tracking accuracy (RMS 0.483, 0.433, 0.511 mm in X,Y,Z) compared to speckle-tracking alone (RMS 0.663, 0.439, 0.613 mm). It also improved the qualitative agreement for color-coded strain meshes to ground truth, and more clearly identified the infarcted region.

Performance optimization of block matching in 3D echocardiography

The introduction of real time 3D echocardiography allows for deformation tracking in three dimensions, without the limitations of 2D methods. However, the processing needs of 3D methods are much higher. The aim of the study was to optimize the performance of 3D block matching by using a Single Instruction Multiple Data (SIMD) model, a technique employed to achieve data level parallelism. Two implementations of SIMD have been tested. The first is based on Streaming SIMD Extensions (SSE), the second uses CUDA, a SIMD architecture proposed by NVIDIA, which is available on several graphics cards. The proposed methods have been validated on synthetic and patient data. With the use of SIMD architecture, the overall processing time is significantly reduced, thus making 3D speckle tracking feasible in a clinical setting. Apart from the implemented sum of square differences (SAD), other matching criteria (e.g. sum of squared differences, normalized cross correlation) can be implemented efficiently (especially on the GPU) thus further improving the accuracy of the block matching process.

Automatic measurement of biparietal diameter with a portable ultrasound device

Knowledge of the exact gestational age and expected day of delivery is essential for providing optimal medical surveillance. Fetal biparietal diameter (BPD) computed by ultrasound has been shown to correlate well with gestational age (before week 22) or it can be used for detecting growth abnormalities, later in the pregnancy. We have been developing a portable ultrasound scanner (the Umoja scanner) that can be used by midwives in LMIC countries with limited ultrasound and technological expertise. The aim of this work was to develop a technique for automatized computation of BPD which can run on tablet devices. An ultrasound image containing a contour of a fetal head is recorded with the prototype scanner. The image is preprocessed: converted to grayscale, gain adjusted, smoothed, dilated/eroded and finally binary thresholded. The potential contour candidates and their Cartesian coordinates are identified by applying the Canny edge detector. A line connecting the two most distant edge contours across the skull is computed. The original grayscale values along this line are used to identify the top and the bottom edge points which are used for measuring the BPD value. All image processing is performed using OpenCV (Open source Computer Vision), which is optimized for tablet devices. 27 ultrasound images suitable for BPD measurement were acquired by an experienced midwife and 9 student midwives with limited or no prior ultrasound experience, on 8 different fetuses from 18 to 34 weeks. Both manual (experienced midwife) and automatic BPD measurements were computed. The correlation plot and the error versus reference plot are produced, the mean error ± 1.96*STD was 0.72±3.62 [mm], while the correlation coefficient was R=0.9932. The automatic measurement failed in 4 cases. The overall computation time on a Nexus 10 tablet was 3.47 seconds, therefore our tool is suited for a portable device with limited computation power. The agreement of the proposed algorithm with the reference measurements is comparable to the interobserver agreement for BPD (2.6 to 3.1 mm from literature study). Testing of the method on an extended dataset, including fetuses at different gestational ages is ongoing.

Fusion of 3D echo and cardiac magnetic resonance volumes during live scanning

The aim of this project was to develop a method for real-time alignment and visualization of 3D echo and CMR volumes during a live 3D echocardiographic scan and to demonstrate its applicability in a typical ultrasound scanning scenario.

A comparison between methods for automatic quantification of global left ventricular function

In recent years, volumetric (3D) cardiac ultrasound imaging has become more readily available in daily clinical practice due to the introduction of matrix array transducer technology. To date, quantitative analysis of these data sets typically requires a significant amount of user interaction. Recently, our teams introduced methods that could help in automating this process. On the one hand, an edge detection algorithm in combination with a deformable subdivision surface was presented for automatic segmentation of the LV cavity. A real-time, dynamic implementation of this segmentation approach in combination with a Kalman filter allows tracking the subendocardial boundary throughout the cardiac cycle. This method is referred to as RCTL. On the other hand, an automatic 3D motion estimation algorithm was presented in which subsequent image volumes are elastically registered using a B-spline transformation field. This method is called splineMIRIT. Both methods were applied to clinical data to extract relevant functional parameters on global left ventricular (LV) function (i.e. stroke volume (SV) and ejection fraction (EF)). Both methods show a good correlation with the reference method and might thus be used for fully automated estimation of global LV function. Given that RCTL is a fully integrated method (accounting for both segmentation and tracking) it seems to be the better approach towards extracting these parameters. However, whether this remains true when assessing parameters for regional LV function remains to be investigated.

A robust Doppler spectral envelope detection technique for automated blood flow measurements

Maximum velocity estimation in a Doppler Spectrogram is of clinical interest. Two existing approaches to maximum velocity estimation, the Geometric method (GM) and the Signal Noise Slope Intersection (SNSI) method are combined in the new approach described in this paper. Further, a data adaptive validity check is introduced that makes the proposed technique independent of user defined parameters. The method enables accurate maximum velocity point detection and obtains an automated, robust tracing of the maximum velocity envelope, which can be further used for automated blood flow measurements. In order to verify the robustness of the algorithm, Pulsatility Index measurements were compared with the observations made by a midwife trained in ultrasound. Evaluation of the steadiness of the location of identified maximum frequency point in simulated flow with varying SNR, ranging to a minimum SNR level of 6dB, also aided the verification. This technique is expected to prove useful in ultrasound machines for diagnoses by clinically inexperienced users.