Spatiotemporal Deep Learning-Based Cine Loop Quality Filter for Handheld Point-of-Care Echocardiography.

The reliability of automated image interpretation of point-of-care (POC) echocardiography scans depends on the quality of the acquired ultrasound data. This work reports on the development and validation of spatiotemporal deep learning models to assess the suitability of input ultrasound cine loops collected using a handheld echocardiography device for processing by an automated quantification algorithm (e.g. ejection fraction estimation). POC echocardiograms (n=885 DICOM cine loops from 175 patients) from two sites were collected using a handheld ultrasound device and annotated for image quality at the frame-level. Attributes of high-quality frames for left ventricular (LV) quantification included a temporally-stable LV, reasonable coverage of LV borders, and good contrast between the borders and chamber. Attributes of low-quality frames included temporal instability of the LV and/or imaging artifacts (e.g., lack of contrast, haze, reverberation, acoustic shadowing). Three different neural network architectures were investigated - (a) frame-level convolutional neural network (CNN) which operates on individual echo frames (VectorCNN), (b) single-stream sequence-level CNN which operates on a sequence of echo frames (VectorCNN+LSTM) and (c) two-stream sequence-level CNNs which operate on a sequence of echo and optical flow frames (VectorCNN+LSTM+Average, VectorCNN+LSTM+MinMax, and VectorCNN+LSTM+ConvPool). Evaluation on a sequestered test dataset containing 76 DICOM cine loops with 16,914 frames showed that VectorCNN+LSTM can effectively utilize both spatial and temporal information to regress the quality of an input frame (accuracy: 0.925, sensitivity = 0.860, specificity = 0.952), compared to the frame-level VectorCNN that only utilizes spatial information in that frame (accuracy: 0.903, sensitivity = 0.791, specificity = 0.949). Furthermore, an independent sample t-test indicated that the cine loops classified to be of adequate quality by the VectorCNN+LSTM model had a statistically significant lower bias in the automatically estimated EF (mean bias = - 3.73 ± 7.46 %, versus a clinically obtained reference EF) compared to the loops classified as inadequate (mean bias = -15.92 ± 12.17 %) (p = 0.007). Thus, cine loop stratification using the proposed spatiotemporal CNN model improves the reliability of automated point-of-care echocardiography image interpretation.

Pragmatic Evaluation of a Deep-Learning Algorithm to Automate Ejection Fraction on Hand-Held, Point-of-Care Echocardiography in a Cardiac Surgical Operating Room.

OBJECTIVE: To test the correlation of ejection fraction (EF) estimated by a deep-learning-based, automated algorithm (Auto EF) versus an EF estimated by Simpson's method. DESIGN: A prospective observational study. SETTING: A single-center study at the Hospital of the University of Pennsylvania. PARTICIPANTS: Study participants were ≥18 years of age and scheduled to undergo valve, aortic, coronary artery bypass graft, heart, or lung transplant surgery. INTERVENTIONS: This noninterventional study involved acquiring apical 4-chamber transthoracic echocardiographic clips using the Philips hand-held ultrasound device, Lumify. MEASUREMENTS AND MAIN RESULTS: In the primary analysis of 54 clips, compared to Simpson's method for EF estimation, bias was similar for Auto EF (-10.17%) and the experienced reader-estimated EF (-9.82%), but the correlation was lower for Auto EF (r = 0.56) than the experienced reader-estimated EF (r = 0.80). In the secondary analyses, the correlation between EF estimated by Simpson's method and Auto EF increased when applied to 27 acquisitions classified as adequate (r = 0.86), but decreased when applied to 27 acquisitions classified as inadequate (r = 0.46). CONCLUSIONS: Applied to acquisitions of adequate image quality, Auto EF produced a numerical EF estimate equivalent to Simpson's method. However, when applied to acquisitions of inadequate image quality, discrepancies arose between EF estimated by Auto EF and Simpson's method. Visual EF estimates by experienced readers correlated highly with Simpson's method in both variable and inadequate imaging conditions, emphasizing its enduring clinical utility.

Using redundancy of round-trip ultrasound signal for non-continuous arrays: Application to gap and blockage compensation.

In ultrasound imaging, an array of elements is used to image a medium. If part of the array is blocked by an obstacle, or if the array is made from several sub-arrays separated by a gap, grating lobes appear and the image is degraded. The grating lobes are caused by missing spatial frequencies, corresponding to the blocked or non-existing elements. However, in an active imaging system, where elements are used both for transmitting and receiving, the round trip signal is redundant: different pairs of transmit and receive elements carry similar information. It is shown here that, if the gaps are smaller than the active sub-apertures, this redundancy can be used to compensate for the missing signals and recover full resolution. Three algorithms are proposed: one is based on a synthetic aperture method, a second one uses dual-apodization beamforming, and the third one is a radio frequency (RF) data based deconvolution. The algorithms are evaluated on simulated and experimental data sets. An application could be imaging through ribs with a large aperture.

A new sensor technology for 2D ultrasound-guided needle tracking.

2D Ultrasound (US) is becoming the preferred modality for image-guided interventions due to its low cost and portability. However, the main limitation is the limited visibility of surgical tools. We present a new sensor technology that can easily be embedded on needles that are used for US-guided interventions. Two different types of materials are proposed to be used as sensor--co-polymer and PZT. The co-polymer technology is particularly attractive due to its plasticity, allowing very thin depositions (10-20 µm) on a variety of needle shapes. Both sensors receive acoustic energy and convert it to an electrical signal. The precise location of the needle can then be estimated from this signal, to provide real-time feedback to the clinician. We evaluated the feasibility of this new technology using (i) a 4DOF robot in a water tank; (ii) extensive ex vivo experiments; and (iii) in vivo studies. Quantitative robotic studies indicated that the co-polymer is more robust and stable when compared to PZT. In quantitative experiments, the technology achieved a tracking accuracy of 0.14 ± 0.03mm, significantly superior to competing technologies. The technology also proved success in near-real clinical studies on tissue data. This sensor technology is non-disruptive of existing clinical workflows, highly accurate, and is cost-effective. Initial clinician feedback shows great potential for large scale clinical impact.

A novel hands-free carotid ultrasound detects low-flow cardiac output in a swine model of pulseless electrical activity arrest.

OBJECTIVE: To determine if a hands-free, noninvasive Doppler ultrasound device can reliably detect low-flow cardiac output by measuring carotid artery blood flow velocities. We compared the ability of observers to detect carotid artery flow velocity differences between pseudo-pulseless electrical activity (PEA) and true-PEA cardiac arrest. METHODS: Five swine were instrumented with aortic (Ao) and right atrial pressure-transducing catheters. The Doppler ultrasound device was adhered to the neck over the carotid artery. Continuous electrocardiogram, pressure readings, and Doppler signal were recorded. Each swine underwent multiple episodes of fibrillation and resuscitation. Episodes of true-PEA and pseudo-PEA were retrospectively identified from all resuscitation attempts by examination of electrocardiogram and Ao waveforms. The sensitivity and specificity of the device to detect pseudo-PEA was obtained using observers blinded to Ao waveform recordings. RESULTS: There was good interobserver reliability related to identification of pseudo- and true-PEA (κ = 0.873). The observers blinded to Ao waveform recordings agreed on 8 of the 9 episodes of pseudo-PEA, whereas 4 false positives of 26 true-PEA events were reported (sensitivity, 0.89; specificity, 0.85). The Doppler device was able to detect carotid flow velocity over a wide range of Ao blood pressures. CONCLUSIONS: This hands-free, noninvasive Doppler ultrasound device can reliably differentiate pseudo-PEA from true-PEA during resuscitation from cardiac arrest, detecting pressure gradient changes of less than 5 mm Hg through to normotension. This device distinguishes conditions of no cardiac output from low cardiac output and may have applications for use during resuscitation from various etiologies of arrest and shock.

Combination of digital mammography with semi-automated 3D breast ultrasound.

This paper describes work aimed at combining 3D ultrasound with full-field digital mammography via a semi-automatic prototype ultrasound scanning mechanism attached to the digital mammography system gantry. Initial efforts to obtain high x-ray and ultrasound image quality through a compression paddle are proving successful. Registration between the x-ray mammogram and ultrasound image volumes is quite promising when the breast is stably compressed. This prototype system takes advantage of many synergies between the co-registered digital mammography and pulse-echo ultrasound image data used for breast cancer detection and diagnosis. In addition, innovative combinations of advanced US and X-ray applications are being implemented and tested along with the basic modes. The basic and advanced applications are those that should provide relatively independent information about the breast tissues. Advanced applications include x-ray tomosynthesis, for 3D delineation of mammographic structures, and non-linear elasticity and 3D color flow imaging by ultrasound, for mechanical and physiological information unavailable from conventional, non-contrast x-ray and ultrasound imaging.

Nonlinear elasticity imaging: theory and phantom study.

In tissue the Young's modulus cannot be assumed constant over a wide deformation range. For example, direct mechanical measurements on human prostate show up to a threefold increase in Young's modulus over a 10% deformation. In conventional elasticity imaging, these effects produce strain-dependent elastic contrast. Ignoring these effects generally leads to suboptimal contrast (stiffer tissues at lower strain are contrasted against softer tissues at higher strain), but measuring the nonlinear behavior results in enhanced tissue differentiation. To demonstrate the methods extracting nonlinear elastic properties, both simulations and measurements were performed on an agar-gelatin phantom. Multiple frames of phase-sensitive ultrasound data are acquired as the phantom is deformed by 12%. All interframe displacement data are brought back to the geometry of the first frame to form a three-dimensional (3-D) data set (depth, lateral, and preload dimensions). Data are fit to a 3-D second order polynomial model for each pixel that adjusts for deformation irregularities. For the phantom geometry and elastic properties considered in this paper, reconstructed frame-to-frame strain images using this model result in improved contrast to noise ratios (CNR) at all preload levels, without any sacrifice in spatial resolution. From the same model, strain hardening at all preload levels can be extracted. This is an independent contrast mechanism. Its maximum CNR occurs at 5.13% preload, and it is a 54% improvement over the best case (preload 10.6%) CNR for frame-to-frame strain reconstruction. Actual phantom measurements confirm the essential features of the simulation. Results show that modeling of the nonlinear elastic behavior has the potential to both increase detectability in elasticity imaging and provide a new independent mechanism for tissue differentiation.

Measuring the nonlinear elastic properties of tissue-like phantoms.

A direct mechanical system simultaneously measuring external force and deformation of samples over a wide dynamic range is used to obtain force-displacement curves of tissue-like phantoms under plain strain deformation. These measurements, covering a wide deformation range, then are used to characterize the nonlinear elastic properties of the phantom materials. The model assumes incompressible media, in which several strain energy potentials are considered. Finite-element analysis is used to evaluate the performance of this material characterization procedure. The procedures developed allow calibration of nonlinear elastic phantoms for elasticity imaging experiments and finite-element simulations.