Delineation of the cardiac structures from 2D echocardiographic images is a common clinical task to establish a diagnosis. Over the past decades, the automation of this task has been the subject of intense research. In this paper, we evaluate how far the state-of-the-art encoder-decoder deep convolutional neural network methods can go at assessing 2D echocardiographic images, i.e. segmenting cardiac structures as well as estimating clinical indices, on a dataset especially designed to answer this objective. We therefore introduce the Cardiac Acquisitions for Multi-structure Ultrasound Segmentation (CA-MUS) dataset, the largest publicly-available and fully-annotated dataset for the purpose of echocardiographic assessment. The dataset contains two and four-chamber acquisitions from 500 patients with reference measurements from one cardiologist on the full dataset and from three cardiologists on a fold of 50 patients. Results show that encoder-decoder based architectures outperform state-of-the-art non-deep learning methods and faithfully reproduce the expert analysis for the end-diastolic and endsystolic left ventricular volumes, with a mean correlation of 0.95 and an absolute mean error of 9.5 ml. Concerning the ejection fraction of the left ventricle, results are more contrasted with a mean correlation coefficient of 0.80 and an absolute mean error of 5.6%. Although these results are below the inter-observer scores, they remain slightly worse than the intra-observer's ones. Based on this observation, areas for improvement are defined, which open the door for accurate and fully-automatic analysis of 2D echocardiographic images.Index Terms-Cardiac segmentation and diagnosis, deep learning, ultrasound, left ventricle, myocardium, left atrium.

Coherent plane wave compounding is a promising technique for achieving very high frame rate imaging without compromising image quality or penetration. However, this approach relies on the hypothesis that the imaged object is not moving during the compounded scan sequence, which is not the case in cardiovascular imaging. This work investigates the effect of tissue motion on retrospective transmit focusing in coherent compounded plane wave imaging (PWI). Two compound scan sequences were studied based on a linear and alternating sequence of tilted plane waves, with different timing characteristics. Simulation studies revealed potentially severe degradations in the retrospective focusing process, where both radial and lateral resolution was reduced, lateral shifts of the imaged medium were introduced, and losses in signal-to-noise ratio (SNR) were inferred. For myocardial imaging, physiological tissue displacements were on the order of half a wavelength, leading to SNR losses up to 35 dB, and reductions of contrast by 40 dB. No significant difference was observed between the different tilt sequences. A motion compensation technique based on cross-correlation was introduced, which significantly recovered the losses in SNR and contrast for physiological tissue velocities. Worst case losses in SNR and contrast were recovered by 35 dB and 27-35 dB, respectively. The effects of motion were demonstrated in vivo when imaging a rat heart. Using PWI, very high frame rates up to 463 fps were achieved at high image quality, but a motion correction scheme was then required.

Proper suppression of tissue clutter is a prerequisite for visualizing flow accurately in ultrasound color flow imaging. Among various clutter suppression methods, the eigen-based filter has shown potential because it can theoretically adapt its stopband to the actual clutter characteristics even when tissue motion is present. This paper presents a formative review on how eigen-based filters should be designed to improve their practical efficacy in adaptively suppressing clutter without affecting the blood flow echoes. Our review is centered around a comparative assessment of two eigen-filter design considerations: 1) eigen-component estimation approach (single-ensemble vs. multi-ensemble formulations), and 2) filter order selection mechanism (eigenvalue-based vs. frequencybased algorithms). To evaluate the practical efficacy of existing eigen-filter designs, we analyzed their clutter suppression level in two in vivo scenarios with substantial tissue motion (intra-operative coronary imaging and thyroid imaging). Our analysis shows that, as compared with polynomial regression filters (with or without instantaneous clutter downmixing), eigen-filters that use a frequency-based algorithm for filter order selection generally give Doppler power images with better contrast between blood and tissue regions. Results also suggest that both multi-ensemble and single-ensemble eigen-estimation approaches have their own advantages and weaknesses in different imaging scenarios. It may be beneficial to develop an algorithmic way of defining the eigen-filter formulation so that its performance advantages can be better realized.

Graft flow is highly dependent on the degree of competitive flow. High competitive flow was found to produce unfavourable WSS consistent with endothelial dysfunction and subsequent graft narrowing and failure. Partial competitive flow, however, may be better tolerated as it was found to be similar to the ideal condition of no competitive flow.