Weakly-supervised learning for catheter segmentation in 3D frustum ultrasound.

Accurate and efficient catheter segmentation in 3D ultrasound (US) is essential for ultrasound-guided cardiac interventions. State-of-the-art segmentation algorithms, based on convolutional neural networks (CNNs), suffer from high computational cost and large 3D data size for GPU implementation, which are far from satisfactory for real-time applications. In this paper, we propose a novel approach for efficient catheter segmentation in 3D US. Instead of using Cartesian US, our approach performs catheter segmentation in Frustum US (i.e., the US data before scan conversion). Compared to Cartesian US, Frustum US has a much smaller volume size, therefore the catheter can be segmented more efficiently in Frustum US. However, annotating the irregular and deformed Frustum images is challenging, and it is laborious to obtain the voxel-level annotation. To address this, we propose a weakly supervised learning framework, which requires only bounding-box annotations. The labels of the voxels are generated by incorporating class activation maps with line filtering, which are iteratively updated during the training cycles. Our experimental results show that, compared to Cartesian US, the catheter can be segmented much more efficiently in Frustum US (i.e., 0.25 s per volume) with better accuracy. Extensive experiments also validate the effectiveness of the proposed weakly supervised learning method.

Medical Instrument Segmentation in 3D US by Hybrid Constrained Semi-Supervised Learning.

Medical instrument segmentation in 3D ultrasound is essential for image-guided intervention. However, to train a successful deep neural network for instrument segmentation, a large number of labeled images are required, which is expensive and time-consuming to obtain. In this article, we propose a semi-supervised learning (SSL) framework for instrument segmentation in 3D US, which requires much less annotation effort than the existing methods. To achieve the SSL learning, a Dual-UNet is proposed to segment the instrument. The Dual-UNet leverages unlabeled data using a novel hybrid loss function, consisting of uncertainty and contextual constraints. Specifically, the uncertainty constraints leverage the uncertainty estimation of the predictions of the UNet, and therefore improve the unlabeled information for SSL training. In addition, contextual constraints exploit the contextual information of the training images, which are used as the complementary information for voxel-wise uncertainty estimation. Extensive experiments on multiple ex-vivo and in-vivo datasets show that our proposed method achieves Dice score of about 68.6%-69.1% and the inference time of about 1 sec. per volume. These results are better than the state-of-the-art SSL methods and the inference time is comparable to the supervised approaches.

Efficient and Robust Instrument Segmentation in 3D Ultrasound Using Patch-of-Interest-FuseNet with Hybrid Loss.

Instrument segmentation plays a vital role in 3D ultrasound (US) guided cardiac intervention. Efficient and accurate segmentation during the operation is highly desired since it can facilitate the operation, reduce the operational complexity, and therefore improve the outcome. Nevertheless, current image-based instrument segmentation methods are not efficient nor accurate enough for clinical usage. Lately, fully convolutional neural networks (FCNs), including 2D and 3D FCNs, have been used in different volumetric segmentation tasks. However, 2D FCN cannot exploit the 3D contextual information in the volumetric data, while 3D FCN requires high computation cost and a large amount of training data. Moreover, with limited computation resources, 3D FCN is commonly applied with a patch-based strategy, which is therefore not efficient for clinical applications. To address these, we propose a POI-FuseNet, which consists of a patch-of-interest (POI) selector and a FuseNet. The POI selector can efficiently select the interested regions containing the instrument, while FuseNet can make use of 2D and 3D FCN features to hierarchically exploit contextual information. Furthermore, we propose a hybrid loss function, which consists of a contextual loss and a class-balanced focal loss, to improve the segmentation performance of the network. With the collected challenging ex-vivo dataset on RF-ablation catheter, our method achieved a Dice score of 70.5%, superior to the state-of-the-art methods. In addition, based on the pre-trained model from ex-vivo dataset, our method can be adapted to the in-vivo dataset on guidewire and achieves a Dice score of 66.5% for a different cardiac operation. More crucially, with POI-based strategy, segmentation efficiency is reduced to around 1.3 seconds per volume, which shows the proposed method is promising for clinical use.

Efficient Medical Instrument Detection in 3D Volumetric Ultrasound Data.

Ultrasound-guided procedures have been applied in many clinical therapies, such as cardiac catheterization and regional anesthesia. Medical instrument detection in 3D Ultrasound (US) is highly desired, but the existing approaches are far from real-time performance. Our objective is to investigate an efficient instrument detection method in 3D US for practical clinical use. We propose a novel Multi-dimensional Mixed Network for efficient instrument detection in 3D US, which extracts the discriminating features at 3D full-image level by a 3D encoder, and then applies a specially designed dimension reduction block to reduce the spatial complexity of the feature maps by projecting from 3D space into 2D space. A 2D decoder is adopted to detect the instrument along the specified axes. By projecting the predicted 2D outputs, the instrument is detected or visualized in the 3D volume. Furthermore, to enable the network to better learn the discriminative information, we propose a multi-level loss function to capture both pixel- and image-level differences. We carried out extensive experiments on two datasets for two tasks: (1) catheter detection for cardiac RF-ablation and (2) needle detection for regional anesthesia. Our experiments show that our proposed method achieves a detection error of 2-3 voxels with an efficiency of about 0.12 sec per 3D US volume. The proposed method is 3-8 times faster than the state-of-the-art methods, leading to real-time performance. The results show that our proposed method has significant clinical value for real-time 3D US-guided intervention.

Doppler Ultrasound Technology for Fetal Heart Rate Monitoring: A Review.

Fetal well-being is commonly assessed by monitoring the fetal heart rate (fHR). In clinical practice, the de facto standard technology for fHR monitoring is based on the Doppler ultrasound (US). Continuous monitoring of the fHR before and during labor is performed using a US transducer fixed on the maternal abdomen. The continuous fHR monitoring, together with simultaneous monitoring of the uterine activity, is referred to as cardiotocography (CTG). In contrast, for intermittent measurements of the fHR, a handheld Doppler US transducer is typically used. In this article, the technology of Doppler US for continuous fHR monitoring and intermittent fHR measurements is described, with emphasis on fHR monitoring for CTG. Special attention is dedicated to the measurement environment, which includes the clinical setting in which fHR monitoring is commonly performed. In addition, to understand the signal content of acquired Doppler US signals, the anatomy and physiology of the fetal heart and the surrounding maternal abdomen are described. The challenges encountered in these measurements have led to different technological strategies, which are presented and critically discussed, with a focus on the US transducer geometry, Doppler signal processing, and fHR extraction methods.

Catheter localization in 3D ultrasound using voxel-of-interest-based ConvNets for cardiac intervention.

PURPOSE: Efficient image-based catheter localization in 3D US during cardiac interventions is highly desired, since it facilitates the operation procedure, reduces the patient risk and improves the outcome. Current image-based catheter localization methods are not efficient or accurate enough for real clinical use. METHODS: We propose a catheter localization method for 3D cardiac ultrasound (US). The catheter candidate voxels are first pre-selected by the Frangi vesselness filter with adaptive thresholding, after which a triplanar-based ConvNet is applied to classify the remaining voxels as catheter or not. We propose a Share-ConvNet for 3D US, which reduces the computation complexity by sharing a single ConvNet for all orthogonal slices. To boost the performance of ConvNet, we also employ two-stage training with weighted cross-entropy. Using the classified voxels, the catheter is localized by a model fitting algorithm. RESULTS: To validate our method, we have collected challenging ex vivo datasets. Extensive experiments show that the proposed method outperforms state-of-the-art methods and can localize the catheter with an average error of 2.1 mm in around 10 s per volume. CONCLUSION: Our method can automatically localize the cardiac catheter in challenging 3D cardiac US images. The efficiency and accuracy localization of the proposed method are considered promising for catheter detection and localization during clinical interventions.

Fetal Heart Rate Monitoring Implemented by Dynamic Adaptation of Transmission Power of a Flexible Ultrasound Transducer Array.

Fetal heart rate (fHR) monitoring using Doppler Ultrasound (US) is a standard method to assess fetal health before and during labor. Typically, an US transducer is positioned on the maternal abdomen and directed towards the fetal heart. Due to fetal movement or displacement of the transducer, the relative fetal heart location (fHL) with respect to the US transducer can change, leading to frequent periods of signal loss. Consequently, frequent repositioning of the US transducer is required, which is a cumbersome task affecting clinical workflow. In this research, a new flexible US transducer array is proposed which allows for measuring the fHR independently of the fHL. In addition, a method for dynamic adaptation of the transmission power of this array is introduced with the aim of reducing the total acoustic dose transmitted to the fetus and the associated power consumption, which is an important requirement for application in an ambulatory setting. The method is evaluated using an in-vitro setup of a beating chicken heart. We demonstrate that the signal quality of the Doppler signal acquired with the proposed method is comparable to that of a standard, clinical US transducer. At the same time, our transducer array is able to measure the fHR for varying fHL while only using 50% of the total transmission power of standard, clinical US transducers.

Catheter segmentation in three-dimensional ultrasound images by feature fusion and model fitting.

Ultrasound (US) has been increasingly used during interventions, such as cardiac catheterization. To accurately identify the catheter inside US images, extra training for physicians and sonographers is needed. As a consequence, automated segmentation of the catheter in US images and optimized presentation viewing to the physician can be beneficial to accelerate the efficiency and safety of interventions and improve their outcome. For cardiac catheterization, a three-dimensional (3-D) US image is potentially attractive because of no radiation modality and richer spatial information. However, due to a limited spatial resolution of 3-D cardiac US and complex anatomical structures inside the heart, image-based catheter segmentation is challenging. We propose a cardiac catheter segmentation method in 3-D US data through image processing techniques. Our method first applies a voxel-based classification through newly designed multiscale and multidefinition features, which provide a robust catheter voxel segmentation in 3-D US. Second, a modified catheter model fitting is applied to segment the curved catheter in 3-D US images. The proposed method is validated with extensive experiments, using different in-vitro, ex-vivo, and in-vivo datasets. The proposed method can segment the catheter within an average tip-point error that is smaller than the catheter diameter (1.9 mm) in the volumetric images. Based on automated catheter segmentation and combined with optimal viewing, physicians do not have to interpret US images and can focus on the procedure itself to improve the quality of cardiac intervention.

Characteristics of Radiofrequency Catheter Ablation Lesion Formation in Real Time In Vivo Using Near Field Ultrasound Imaging.

OBJECTIVES: Visualizing myocardium with near field ultrasound (NFUS) transducers in the tip of the catheter might provide an image of the evolving pathological lesion during energy delivery. BACKGROUND: Radiofrequency (RF) catheter ablation has been effective in arrhythmia treatment, but no technology has allowed lesion formation to be visualized in real time in vivo. METHODS: RF catheter ablations were performed in vivo with the goal to create transmural atrial lesions and large ventricular lesions. RF lesion formation was imaged in real time using M-mode, tissue Doppler, and strain rate information from the NFUS open irrigated RF ablation catheter incorporating 4 ultrasound transducers (1 axial and 3 radial), and growth kinetics were analyzed. Nineteen dogs underwent ablation in the right and left atria (n = 185), right ventricle (n = 67), and left ventricle (n = 66). Lesions were echolucent with tissue strain rate by NFUS. RESULTS: Lesion growth frequently progressed from epicardium to endocardium in thin-walled tissue. The half time of lesion growth was 5.5 ± 2.8 s in thin-walled and 9.7 ± 4.3 s in thick-walled tissue. Latency of lesion onset was seen in 57% of lesions ranging from 1 to 63.8 s. Tissue edema (median 25% increased wall thickness) formed immediately upon lesion formation in 83%, and intramyocardial steam was seen in 71% of cases. CONCLUSIONS: NFUS was effective in imaging RF catheter ablation lesion formation in real time. It was useful in assessing the dynamics of lesion growth and could visualize impending steam pops. It may be a useful technology to improve both safety and efficacy of RF catheter ablation.

Quantitative imaging performance of frequency-tunable capacitive micromachined ultrasonic transducer array designed for intracardiac application: Phantom study.

Commercially available intracardiac echo (ICE) catheters face a trade-off between viewing depth and resolution. Frequency-tunable ICE probes would offer versatility of choice between penetration or resolution imaging within a single device. In this phantom study, the imaging performance of a novel, frequency-tunable, 32-element, 1-D CMUT array integrated with front-end electronics is evaluated. Phased-array ultrasound imaging with a forward-looking CMUT probe prototype operated beyond collapse mode at voltages up to three times higher than the collapse voltage (-65 V) is demonstrated. Imaging performance as a function of bias voltage (-70 V to -160 V), transmit pulse frequency (5-25 MHz), and number of transmit pulse cycles (1-3) is quantified, based on which penetration, resolution, and generic imaging modes are identified. It is shown that by utilizing the concept of frequency tuning, images with different characteristics can be generated trading-off the resolution and penetration depth. The penetration mode provides imaging up to 71 mm in the tissue-mimicking phantom, axial resolution of 0.44 mm, and lateral resolution of 0.12 rad. In the resolution mode, axial resolution of 0.055 mm, lateral resolution of 0.035 rad, and penetration depth of 16 mm are measured. These results show what this CMUT array has the potential versatile characteristics needed for intracardiac imaging, despite its relatively small transducer aperture size of 2 mm × 2 mm imposed by the clinical application.

Near-Field Ultrasound Imaging During Radiofrequency Catheter Ablation: Tissue Thickness and Epicardial Wall Visualization and Assessment of Radiofrequency Ablation Lesion Formation and Depth.

BACKGROUND: Safe and successful radiofrequency catheter ablation depends on creation of transmural lesions without collateral injury to contiguous structures. Near-field ultrasound (NFUS) imaging through transducers in the tip of an ablation catheter may provide important information about catheter contact, wall thickness, and ablation lesion formation. METHODS AND RESULTS: NFUS imaging was performed using a specially designed open-irrigated radiofrequency ablation catheter incorporating 4 ultrasound transducers. Tissue/phantom thickness was measured in vitro with varying contact angles. In vivo testing was performed in 19 dogs with NFUS catheters positioned in 4 chambers. Wall thickness measurements were made at 222 sites (excluding the left ventricle) and compared with measurements from intracardiac echocardiography. Imaging was used to identify the epicardium with saline infusion into the pericardial space at 39 sites. In vitro, the measured exceeded actual tissue/phantom thickness by 13% to 20%. In vivo, NFUS reliably visualized electrode-tissue contact, but sensitivity of epicardial imaging was 92%. The chamber wall thickness measured by NFUS correlated well with intracardiac echocardiography (r=0.86; P<0.0001). Sensitivity of lesion identification by NFUS was 94% for atrial and 95% for ventricular ablations. NFUS was the best parameter to predict lesion depth in right and left ventricle (r=0.47; P<0.0001; multiple regression P=0.0025). Lesion transmurality was correctly identified in 87% of atrial lesions. CONCLUSIONS: NFUS catheter imaging reliably assesses electrode-tissue contact and wall thickness. Its use during radiofrequency catheter ablation may allow the operator to assess the depth of ablation required for transmural lesion formation to optimize power delivery.

Improved ultrasound transducer positioning by fetal heart location estimation during Doppler based heart rate measurements.

OBJECTIVE: Doppler ultrasound (US) is the most commonly applied method to measure the fetal heart rate (fHR). When the fetal heart is not properly located within the ultrasonic beam, fHR measurements often fail. As a consequence, clinical staff need to reposition the US transducer on the maternal abdomen, which can be a time consuming and tedious task. APPROACH: In this article, a method is presented to aid clinicians with the positioning of the US transducer to produce robust fHR measurements. A maximum likelihood estimation (MLE) algorithm is developed, which provides information on fetal heart location using the power of the Doppler signals received in the individual elements of a standard US transducer for fHR recordings. The performance of the algorithm is evaluated with simulations and in vitro experiments performed on a beating-heart setup. MAIN RESULTS: Both the experiments and the simulations show that the heart location can be accurately determined with an error of less than 7 mm within the measurement volume of the employed US transducer. SIGNIFICANCE: The results show that the developed algorithm can be used to provide accurate feedback on fetal heart location for improved positioning of the US transducer, which may lead to improved measurements of the fHR.

Preclinical Testing of Frequency-Tunable Capacitive Micromachined Ultrasonic Transducer Probe Prototypes.

In intracardiac echocardiography (ICE) it may be beneficial to generate ultrasound images acquired at multiple frequencies, having the possibility of high penetration or high-resolution imaging in a single device. The objective of the presented work is to test two frequency-tunable probe prototypes in a preclinical setting: a rigid probe having a diameter of 11 mm and a new flexible and steerable 12-Fr ICE catheter. Both probes feature a forward-looking 32-element capacitive micromachined ultrasonic transducer array (aperture of 2 × 2 mm2) operated in collapse mode, which allows for frequency tuning in the 6-MHz-18-MHz range. The rigid probe prototype is tested ex vivo in a passive heart platform. Images of an aortic valve acquired in high-penetration (6 MHz), generic (12 MHz) and high-resolution (18 MHz) mode combine satisfying image quality and penetration depth between 2.5 cm and 10 cm. The ICE catheter prototype is tested in vivo using a porcine animal model. Images of an aortic valve are acquired in the 3 imaging modes with the ICE catheter placed in an ascending aorta at multiple depths. It was found that the combination of the forward-looking design and frequency-tuning capability allows visualizing intracardiac structures of various sizes at different distances relative to the catheter tip, providing both wide overviews and detailed close-ups.

Medical Instrument Detection in 3-Dimensional Ultrasound Data Volumes.

Ultrasound-guided medical interventions are broadly applied in diagnostics and therapy, e.g., regional anesthesia or ablation. A guided intervention using 2-D ultrasound is challenging due to the poor instrument visibility, limited field of view, and the multi-fold coordination of the medical instrument and ultrasound plane. Recent 3-D ultrasound transducers can improve the quality of the image-guided intervention if an automated detection of the needle is used. In this paper, we present a novel method for detecting medical instruments in 3-D ultrasound data that is solely based on image processing techniques and validated on various ex vivo and in vivo data sets. In the proposed procedure, the physician is placing the 3-D transducer at the desired position, and the image processing will automatically detect the best instrument view, so that the physician can entirely focus on the intervention. Our method is based on the classification of instrument voxels using volumetric structure directions and robust approximation of the primary tool axis. A novel normalization method is proposed for the shape and intensity consistency of instruments to improve the detection. Moreover, a novel 3-D Gabor wavelet transformation is introduced and optimally designed for revealing the instrument voxels in the volume, while remaining generic to several medical instruments and transducer types. Experiments on diverse data sets, including in vivo data from patients, show that for a given transducer and an instrument type, high detection accuracies are achieved with position errors smaller than the instrument diameter in the 0.5-1.5-mm range on average.

Visualizing intramyocardial steam formation with a radiofrequency ablation catheter incorporating near-field ultrasound.

INTRODUCTION: Steam pops are a risk of irrigated RF ablation even when limiting power delivery. There is currently no way to predict gas formation during ablation. It would be useful to visualize intramyocardial gas formation prior to a steam pop occurring using near-field ultrasound integrated into a RF ablation catheter. METHODS AND RESULTS: In an in vivo open-chest ovine model (n = 9), 86 lesions were delivered to the epicardial surface of the ventricles. Energy was delivered for 15-60 seconds, to achieve lesions with and without steam pops, based on modeling data. The ultrasound image was compared to a digital audio recording from within the pericardium by a blinded observer. Of 86 lesions, 28 resulted in an audible steam pop. For lesions that resulted in a steam pop compared to those that did not (n = 58), the mean power delivered was 8.0 ± 1.8 W versus 6.7 ± 2.0 W, P = 0.006. A change in US contrast due to gas formation in the tissue occurred in all lesions that resulted in a steam pop. In 4 ablations, a similar change in US contrast was observed in the tissue and RF delivery was stopped; in these cases, no pop occurred. The mean depth of gas formation was 0.9 ± 0.8 mm, which correlated with maximal temperature predicted by modeling. Changes in US contrast occurred 7.6 ± 7.2 seconds before the impedance rise and 7.9 ± 6.2 seconds (0.1-17.0) before an audible pop. CONCLUSION: Integrated US in an RF ablation catheter is able to visualize gas formation intramyocardially several seconds prior to a steam pop occurring. This technology may help prevent complications arising from steam pops.

Effects of spatially targeted transcutaneous electrical nerve stimulation using an electrode array that measures skin resistance on pain and mobility in patients with osteoarthritis in the knee: a randomized controlled trial.

A novel device was developed that measured local electrical skin resistance and generated pulsed local electrical currents that were delivered across the skin around the knee for patients with osteoarthritis (termed eBrace TENS). Currents were delivered using an electrode array of 16 small circular electrode elements so that stimulation could be spatially targeted. The aim of this study was to investigate the effects of spatially targeted transcutaneous electrical nerve stimulation (TENS) to points of low skin resistance on pain relief and mobility in osteoarthritis of the knee (OAK). A randomised, controlled, 3-arm, parallel-group trial was designed that compared pain and function following a 30 to 45 minute intervention of TENS at specific locations depending on the local electrical skin resistance. Pain intensity by the visual analogue scale (VAS), 6-minute walk test, maximum voluntary contraction (MVC), and range-of-motion (ROM) were the primary outcomes. Lowest-resistance TENS reduced pain intensity during walking relative to resting baseline compared with random TENS (95% confidence interval of the difference: -20.8mm, -1.26 mm). There were no statistically significant differences between groups in distance during the walk test, maximum voluntary contraction (MVC) or range-of-motion (ROM) measures or WOMAC scores. In conclusion, we provide evidence that use of a matrix electrode that spatially targets strong nonpainful TENS for 30 to 45 minutes at sites of low resistance can reduce pain intensity at rest and during walking.

Characterization of cardiovascular liver motion for the eventual application of elasticity imaging to the liver in vivo.

Elastography, which uses ultrasound to image the tissue strain that results from an applied displacement, can display tumours and heat-ablated tissue with high contrast. However, its application to liver in vivo may be problematic due to the presence of respiratory and cardiovascular sources of displacement. The aim of this study was to measure the cardiovascular-induced component of natural liver motion for the purpose of planning future work that will either use the motion to produce elasticity images or will compensate for it when employing an external source of displacement. A total of 36 sequences of 7 s real-time radio frequency (RF) echo images of the liver were acquired from six healthy volunteers during breath-hold using a stationary 3.5 MHz transducer. For each image sequence, the axial and lateral components of displacement were measured for each pair of consecutive RF images using 2D-echo tracking. The spatio-temporal character of these displacements was then analysed using a novel approach, employing proper orthogonal decomposition, whereby the dominant motion patterns are described by eigenvectors with the highest eigenvalues. The motion patterns of different liver segments were complex, but they were also found to be cyclic, highly repeatable and capable of producing measurable displacements in the liver. These observations provide good evidence to suggest that it may be possible to correct for natural liver motion when using an externally applied displacement for elasticity imaging. It was also found that about 65%-70% of all liver motion could be described using the first eigenvector. Use of only this component of the motion will greatly simplify the design of a mechanical system to be used in an objective study of elasticity imaging of phantoms and excised tissues in the presence of simulated cardiovascular-induced liver motion.