Multi-Task Learning for Motion Analysis and Segmentation in 3D Echocardiography.

Characterizing left ventricular deformation and strain using 3D+time echocardiography provides useful insights into cardiac function and can be used to detect and localize myocardial injury. To achieve this, it is imperative to obtain accurate motion estimates of the left ventricle. In many strain analysis pipelines, this step is often accompanied by a separate segmentation step; however, recent works have shown both tasks to be highly related and can be complementary when optimized jointly. In this work, we present a multi-task learning network that can simultaneously segment the left ventricle and track its motion between multiple time frames. Two task-specific networks are trained using a composite loss function. Cross-stitch units combine the activations of these networks by learning shared representations between the tasks at different levels. We also propose a novel shape-consistency unit that encourages motion propagated segmentations to match directly predicted segmentations. Using a combined synthetic and in-vivo 3D echocardiography dataset, we demonstrate that our proposed model can achieve excellent estimates of left ventricular motion displacement and myocardial segmentation. Additionally, we observe strong correlation of our image-based strain measurements with crystal-based strain measurements as well as good correspondence with SPECT perfusion mappings. Finally, we demonstrate the clinical utility of the segmentation masks in estimating ejection fraction and sphericity indices that correspond well with benchmark measurements.

Delayed cancer diagnosis in the pregnant patient: navigating a complex medical and ethical dilemma.

Prompt diagnosis of cancer in pregnancy is necessary to ensure timely management and improve outcomes. However, there are a several reasons why diagnosis may be delayed in pregnancy. Three major contributors to delayed diagnosis and treatment are patient delay, provider delay, and referral delay. This article aims to (1) increase physician awareness of this problem by providing a detailed review of the main culprits of delayed diagnosis and treatment of cancer in the pregnant patient, (2) discuss the complex ethical issues at hand in these cases, and (3) provide suggestions on how to better address such cases with the goal of improving patient outcomes.

Melanoma in pregnancy.

Melanoma is one of the most common types of cancer diagnosed during pregnancy. Patients with advanced disease require frequent staging examinations (e.g., CT, PET, MRI, ultrasound), which, during pregnancy must be modified from routine protocol to minimize risk to the fetus. We will review the diagnostic and treatment approach to pregnant patients with melanoma, with a discussion and pictorial examples of imaging protocol modifications, and the appearance of metastatic melanoma on radiology exams using modified protocols due to pregnancy.

Thyroid cancer in pregnancy: diagnosis, management, and treatment.

The evaluation and management of cancer during pregnancy requires special care to assure the health and safety of both the mother and fetus. The diagnosis and treatment of thyroid cancer in the non-pregnant patient often involves radioactive iodine exposure. However, radioactive iodine is contraindicated in pregnancy and surgical interventions pose risks to both the mother and fetus. Thus, the management of thyroid cancer during pregnancy is a unique clinical challenge. In this review, we discuss the imaging of thyroid nodules during pregnancy, including the role of CT, MRI, and nuclear Imaging, as well as that of Ultrasound and FNA. The staging and prognosis are discussed along with the management, treatment, and surveillance of thyroid cancer in pregnancy. Finally, the risks to the fetus through treatment are examined. Case examples are provided with an emphasis on the appropriate direction of care from a radiologist's perspective.

Co-attention spatial transformer network for unsupervised motion tracking and cardiac strain analysis in 3D echocardiography.

Myocardial ischemia/infarction causes wall-motion abnormalities in the left ventricle. Therefore, reliable motion estimation and strain analysis using 3D+time echocardiography for localization and characterization of myocardial injury is valuable for early detection and targeted interventions. Previous unsupervised cardiac motion tracking methods rely on heavily-weighted regularization functions to smooth out the noisy displacement fields in echocardiography. In this work, we present a Co-Attention Spatial Transformer Network (STN) for improved motion tracking and strain analysis in 3D echocardiography. Co-Attention STN aims to extract inter-frame dependent features between frames to improve the motion tracking in otherwise noisy 3D echocardiography images. We also propose a novel temporal constraint to further regularize the motion field to produce smooth and realistic cardiac displacement paths over time without prior assumptions on cardiac motion. Our experimental results on both synthetic and in vivo 3D echocardiography datasets demonstrate that our Co-Attention STN provides superior performance compared to existing methods. Strain analysis from Co-Attention STNs also correspond well with the matched SPECT perfusion maps, demonstrating the clinical utility for using 3D echocardiography for infarct localization.

Simultaneous Segmentation and Motion Estimation of Left Ventricular Myocardium in 3D Echocardiography Using Multi-task Learning.

Motion estimation and segmentation are both critical steps in identifying and assessing myocardial dysfunction, but are traditionally treated as unique tasks and solved as separate steps. However, many motion estimation techniques rely on accurate segmentations. It has been demonstrated in the computer vision and medical image analysis literature that both these tasks may be mutually beneficial when solved simultaneously. In this work, we propose a multi-task learning network that can concurrently predict volumetric segmentations of the left ventricle and estimate motion between 3D echocardiographic image pairs. The model exploits complementary latent features between the two tasks using a shared feature encoder with task-specific decoding branches. Anatomically inspired constraints are incorporated to enforce realistic motion patterns. We evaluate our proposed model on an in vivo 3D echocardiographic canine dataset. Results suggest that coupling these two tasks in a learning framework performs favorably when compared against single task learning and other alternative methods.

Diagnostic Accuracy of Shear-Wave Elastography for Breast Lesion Characterization in Women: A Systematic Review and Meta-Analysis.

PURPOSE: The aim of this study was to assess the diagnostic accuracy of 2-D shear-wave elastography (SWE) for differentiating benign and malignant breast lesions in women with abnormal findings on mammography. METHODS: Included in this review are studies of diagnostic accuracy published before June 2021 using 2-D SWE to evaluate female breast lesions. Included studies were required to include at least 50 lesions, report quantitative shear-wave speed (SWS) thresholds, and include a reference standard of either biopsy or 2-year stability. Included studies used the mean, maximum, minimum, or SD of SWS for classification. A systematic search of PubMed, Scopus, Embase, Ovid-MEDLINE, the Cochrane Library, and Web of Science was performed. Bias and applicability of the studies were assessed using Quality Assessment of Diagnostic Accuracy Studies 2. A hierarchical summary receiver operating characteristic model was used to arrive at the summary statistics. RESULTS: Eighty-seven prospective and retrospective studies were included, encompassing 17,810 women (mean age 42.3 ± 10.4 years) with 19,043 lesions (7,623 malignant). Summary sensitivities and specificities, respectively, were 0.86 (95% confidence interval [CI], 0.83-0.88) and 0.87 (95% CI, 0.84-0.88) for mean SWS, 0.83 (95% CI, 0.80-0.85) and 0.88 (95% CI, 0.86-0.90) for the maximum, 0.86 (95% CI, 0.74-0.93) and 0.81 (95% CI, 0.69-0.89) for the minimum, and 0.82 (95% CI, 0.77-0.86) and 0.88 (95% CI, 0.85-0.91) for the SD. Alternatively, the areas under the receiver operating characteristic curve were 0.93 (95% CI, 0.91-0.94), 0.92 (95% CI, 0.90-0.94), 0.90 (95% CI, 0.82-0.96), and 0.92 (95% CI, 0.88-0.94), respectively. CONCLUSIONS: This review demonstrates the discriminative power of SWE in the diagnosis of breast cancer. Using the resulting likelihood ratios, SWE may prove beneficial in downgrading BI-RADS® 4a or upgrading BI-RADS 3 lesions. However, current society guidelines do not provide definitive recommendations regarding the use of SWE and its counterpart strain elastography (SE). Comparison with our results suggests that SE alone or a combination of SE and SWE may provide better diagnostic performance than SWE alone and serve as an adjunct to current diagnostic techniques.

Wunderlich Syndrome: Wonder What It Is.

Wunderlich syndrome (WS) refers to spontaneous renal or perinephric hemorrhage occurring in the absence of known trauma. WS is much less common than hemorrhage occurring after iatrogenic or traumatic conditions. Lenk's triad of acute onset flank pain, flank mass, and hypovolemic shock is a classic presentation of WS but seen in less than a quarter of patients. The majority of patients present only with isolated flank pain and often imaged with an unenhanced CT in the emergency department. The underlying etiology is varied with most cases attributed to neoplasms, vascular disease, cystic renal disease and anticoagulation induced; the etiology is often occult on the initial exam and further evaluation is necessary. Urologists are familiar with this unique entity but radiologists, who are more likely to be the first to diagnose WS, may not be familiar with the imaging work up and management options. In the last decade or so, there has been a conspicuous shift in the approach to WS and thus it will be worthwhile to revisit WS in detail. In our review, we will review the multimodality imaging approach to WS, describe optimal follow up and elaborate on management.

Multi-frame Attention Network for Left Ventricle Segmentation in 3D Echocardiography.

Echocardiography is one of the main imaging modalities used to assess the cardiovascular health of patients. Among the many analyses performed on echocardiography, segmentation of left ventricle is crucial to quantify the clinical measurements like ejection fraction. However, segmentation of left ventricle in 3D echocardiography remains a challenging and tedious task. In this paper, we propose a multi-frame attention network to improve the performance of segmentation of left ventricle in 3D echocardiography. The multi-frame attention mechanism allows highly correlated spatiotemporal features in a sequence of images that come after a target image to be used to augment the performance of segmentation. Experimental results shown on 51 in vivo porcine 3D+time echocardiography images show that utilizing correlated spatiotemporal features significantly improves the performance of left ventricle segmentation when compared to other standard deep learning-based medical image segmentation models.

Endoscopic Retrograde Cholangiopancreatography: Deciphering the Black and White.

Endoscopic Retrograde Cholangiopancreatography (ERCP) remains the conventional method of imaging the pancreatic and biliary tree and is performed by direct injection of iodinated contrast material via the major papilla. This diagnostic procedure gained popularity in the 1970s and subsequently paved way for ERCP guided interventions such as sphincterotomy, stone retrieval and stent placement. Currently, therapeutic ERCP is more widespread than diagnostic ERCP primarily due to the availability of noninvasive imaging. Nevertheless, more than half a million ERCPs are performed annually in the United States and radiologists need to be comfortable interpreting them. The following review will familiarize the reader with the imaging appearances of biliary and pancreatic disorders on conventional ERCP, and elaborate on therapeutic ERCP with illustrative examples.

Synthesis of fracture radiographs with deep neural networks.

PURPOSE: We describe a machine learning system for converting diagrams of fractures into realistic X-ray images. We further present a method for iterative, human-guided refinement of the generated images and show that the resulting synthetic images can be used during training to increase the accuracy of deep classifiers on clinically meaningful subsets of fracture X-rays. METHODS: A neural network was trained to reconstruct images from programmatically created line drawings of those images. The images were then further refined with an optimization-based technique. Ten physicians were recruited into a study to assess the realism of synthetic radiographs created by the neural network. They were presented with mixed sets of real and synthetic images and asked to identify which images were synthetic. Two classifiers were trained to detect humeral shaft fractures: one only on true fracture images, and one on both true and synthetic images. RESULTS: Physicians were 49.63% accurate in identifying whether images were synthetic or real. This is close to what would be expected by pure chance (i.e. random guessing). A classifier trained only on real images detected fractures with 67.21% sensitivity when no fracture fixation hardware was present. A classifier trained on both real images and synthetic images was 75.54% sensitive. CONCLUSION: Our method generates X-rays realistic enough to be indistinguishable from real X-rays. We also show that synthetic images generated using this method can be used to increase the accuracy of deep classifiers on clinically meaningful subsets of fracture X-rays.

Measurement of Liver Stiffness Using Shear Wave Elastography in a Rat Model: Factors Impacting Stiffness Measurement with Multiple- and Single-Tracking-Location Techniques.

The clinical use of elastography for monitoring fibrosis progression is challenged by the subtle changes in liver stiffness associated with early-stage fibrosis and the comparatively large variance in stiffness estimates provided by elastography. Single-tracking-location (STL) shear wave elasticity imaging (SWEI) is an ultrasound elastography technique previously found to provide improved estimate precision compared with multiple-tracking-location (MTL) SWEI. Because of the improved precision, it is reasonable to expect that STL-SWEI would provide improved ability to differentiate liver fibrosis stage compared with MTL-SWEI. However, this expectation has not been previously challenged rigorously. In this work, the performance of STL- and MTL-SWEI in the setting of a rat model of liver fibrosis is characterized, and the advantages of STL-SWEI in staging fibrosis are explored. The purpose of this study was to determine what advantages, if any, arise from using STL-SWEI instead of MTL-SWEI in the characterization of fibrotic liver. Thus, the ability of STL-SWEI to differentiate livers at various METAVIR fibrosis scores, for ex vivo postmortem measurements, is explored. In addition, we examined the effect of the common confounding factor of fluid versus solid boundary conditions in SWEI experiments. Sprague-Dawley rats were treated with carbon tetrachloride over several weeks to produce liver disease of varying severity. STL and MTL stiffness measurements were performed ex vivo and compared with the METAVIR scores from histological analysis and the duration of treatment. A strong association was observed between liver stiffness and weeks of treatment with the liver toxin carbon tetrachloride. Direct comparison of STL- and MTL-SWEI measurements revealed no significant difference in ability to differentiate fibrosis stages based on SWEI mean values. However, image interquartile range was greatly improved in the case of STL-SWEI, compared with MTL-SWEI, at small beam spacing.

Shear wave arrival time estimates correlate with local speckle pattern.

We present simulation and phantom studies demonstrating a strong correlation between errors in shear wave arrival time estimates and the lateral position of the local speckle pattern in targets with fully developed speckle. We hypothesize that the observed arrival time variations are largely due to the underlying speckle pattern, and call the effect speckle bias. Arrival time estimation is a key step in quantitative shear wave elastography, performed by tracking tissue motion via cross-correlation of RF ultrasound echoes or similar methods. Variations in scatterer strength and interference of echoes from scatterers within the tracking beam result in an echo that does not necessarily describe the average motion within the beam, but one favoring areas of constructive interference and strong scattering. A swept-receive image, formed by fixing the transmit beam and sweeping the receive aperture over the region of interest, is used to estimate the local speckle pattern. Metrics for the lateral position of the speckle are found to correlate strongly (r > 0.7) with the estimated shear wave arrival times both in simulations and in phantoms. Lateral weighting of the swept-receive pattern improved the correlation between arrival time estimates and speckle position. The simulations indicate that high RF echo correlation does not equate to an accurate shear wave arrival time estimate-a high correlation coefficient indicates that motion is being tracked with high precision, but the location tracked is uncertain within the tracking beam width. The presence of a strong on-axis speckle is seen to imply high RF correlation and low bias. The converse does not appear to be true-highly correlated RF echoes can still produce biased arrival time estimates. The shear wave arrival time bias is relatively stable with variations in shear wave amplitude and sign (-20 µm to 20 µm simulated) compared with the variation with different speckle realizations obtained along a given tracking vector. We show that the arrival time bias is weakly dependent on shear wave amplitude compared with the variation with axial position/ local speckle pattern. Apertures of f/3 to f/8 on transmit and f/2 and f/4 on receive were simulated. Arrival time error and correlation with speckle pattern are most strongly determined by the receive aperture.

Scholte wave generation during single tracking location shear wave elasticity imaging of engineered tissues.

The physical environment of engineered tissues can influence cellular functions that are important for tissue regeneration. Thus, there is a critical need for noninvasive technologies capable of monitoring mechanical properties of engineered tissues during fabrication and development. This work investigates the feasibility of using single tracking location shear wave elasticity imaging (STL-SWEI) for quantifying the shear moduli of tissue-mimicking phantoms and engineered tissues in tissue engineering environments. Scholte surface waves were observed when STL-SWEI was performed through a fluid standoff, and confounded shear moduli estimates leading to an underestimation of moduli in regions near the fluid-tissue interface.

Single tracking location acoustic radiation force impulse viscoelasticity estimation (STL-VE): A method for measuring tissue viscoelastic parameters.

Single tracking location (STL) shear wave elasticity imaging (SWEI) is a method for detecting elastic differences between tissues. It has the advantage of intrinsic speckle bias suppression compared with multiple tracking location variants of SWEI. However, the assumption of a linear model leads to an overestimation of the shear modulus in viscoelastic media. A new reconstruction technique denoted single tracking location viscosity estimation (STL-VE) is introduced to correct for this overestimation. This technique utilizes the same raw data generated in STL-SWEI imaging. Here, the STL-VE technique is developed by way of a maximum likelihood estimation for general viscoelastic materials. The method is then implemented for the particular case of the Kelvin-Voigt Model. Using simulation data, the STL-VE technique is demonstrated and the performance of the estimator is characterized. Finally, the STL-VE method is used to estimate the viscoelastic parameters of ex vivo bovine liver. We find good agreement between the STL-VE results and the simulation parameters as well as between the liver shear wave data and the modeled data fit.

Minimization of displacement estimation bias due to high amplitude-reflections using envelope-weighted normalization.

In elastography, displacement estimation is often performed using cross-correlation-based techniques, assuming fully-developed, homogeneous speckle. In the presence of a local, large variation in echo amplitude, such as a reflection from a vessel wall, this assumption does not hold true, resulting in a biased displacement estimate. Normalizing the echo by its envelope before displacement estimation reduces this effect at the cost of larger jitter errors. An algorithm is proposed to reduce amplitude-dependent bias in displacement estimates while avoiding a large increase in the jitter error magnitude. The algorithm involves 'Envelope-Weighted Normalization' (EWN) of echo data before displacement estimation. A parametric analysis was conducted to find the optimum parameters with which this technique could be implemented. The EWN technique was found to significantly reduce the rms error of the displacement estimates, showing the greatest improvements when utilizing longer window lengths and higher ultrasonic frequencies.