Automated classification of liver fibrosis stages using ultrasound imaging.

BACKGROUND: Ultrasound imaging is the most frequently performed for the patients with chronic hepatitis or liver cirrhosis. However, ultrasound imaging is highly operator dependent and interpretation of ultrasound images is subjective, thus well-trained radiologist is required for evaluation. Automated classification of liver fibrosis could alleviate the shortage of skilled radiologist especially in low-to-middle income countries. The purposed of this study is to evaluate deep convolutional neural networks (DCNNs) for classifying the degree of liver fibrosis according to the METAVIR score using US images. METHODS: We used ultrasound (US) images from two tertiary university hospitals. A total of 7920 US images from 933 patients were used for training/validation of DCNNs. All patient were underwent liver biopsy or hepatectomy, and liver fibrosis was categorized based on pathology results using the METAVIR score. Five well-established DCNNs (VGGNet, ResNet, DenseNet, EfficientNet and ViT) was implemented to predict the METAVIR score. The performance of DCNNs for five-level (F0/F1/F2/F3/F4) classification was evaluated through area under the receiver operating characteristic curve (AUC) with 95% confidential interval, accuracy, sensitivity, specificity, positive and negative likelihood ratio. RESULTS: Similar mean AUC values were achieved for five models; VGGNet (0.96), ResNet (0.96), DenseNet (0.95), EfficientNet (0.96), and ViT (0.95). The same mean accuracy (0.94) and specificity values (0.96) were yielded for all models. In terms of sensitivity, EffcientNet achieved highest mean value (0.85) while the other models produced slightly lower values range from 0.82 to 0.84. CONCLUSION: In this study, we demonstrated that DCNNs can classify the staging of liver fibrosis according to METAVIR score with high performance using conventional B-mode images. Among them, EfficientNET that have fewer parameters and computation cost produced highest performance. From the results, we believe that DCNNs based classification of liver fibrosis may allow fast and accurate diagnosis of liver fibrosis without needs of additional equipment for add-on test and may be powerful tool for supporting radiologists in clinical practice.

Reverse Scan Conversion and Efficient Deep Learning Network Architecture for Ultrasound Imaging on a Mobile Device.

Point-of-care ultrasound (POCUS), realized by recent developments in portable ultrasound imaging systems for prompt diagnosis and treatment, has become a major tool in accidents or emergencies. Concomitantly, the number of untrained/unskilled staff not familiar with the operation of the ultrasound system for diagnosis is increasing. By providing an imaging guide to assist clinical decisions and support diagnosis, the risk brought by inexperienced users can be managed. Recently, deep learning has been employed to guide users in ultrasound scanning and diagnosis. However, in a cloud-based ultrasonic artificial intelligence system, the use of POCUS is limited due to information security, network integrity, and significant energy consumption. To address this, we propose (1) a structure that simultaneously provides ultrasound imaging and a mobile device-based ultrasound image guide using deep learning, and (2) a reverse scan conversion (RSC) method for building an ultrasound training dataset to increase the accuracy of the deep learning model. Experimental results show that the proposed structure can achieve ultrasound imaging and deep learning simultaneously at a maximum rate of 42.9 frames per second, and that the RSC method improves the image classification accuracy by more than 3%.

In Vivo Evaluation of Plane Wave Imaging for Abdominal Ultrasonography.

Although plane wave imaging (PWI) has been extensively employed for ultrafast ultrasound imaging, its potential for sectorial B-mode imaging with a convex array transducer has not yet been widely recognized. Recently, we reported an optimized PWI approach for sector scanning that exploits the dynamic transmit focusing capability. In this paper, we first report the clinical applicability of the optimized PWI for abdominal ultrasonography by in vivo image and video evaluations and compare it with conventional focusing (CF) and diverging wave imaging (DWI), which is another dynamic transmit focusing technique generally used for sectorial imaging. In vivo images and videos of the liver, kidney, and gallbladder were obtained from 30 healthy volunteers using PWI, DWI, and CF. Three radiologists assessed the phantom images, 156 in vivo images, and 66 in vivo videos. PWI showed significantly enhanced (p < 0.05) spatial resolution, contrast, and noise and artifact reduction, and a 4-fold higher acquisition rate compared to CF and provided similar performances compared to DWI. Because the computations required for PWI are considerably lower than that for DWI, PWI may represent a promising technique for sectorial imaging in abdominal ultrasonography that provides better image quality and eliminates the need for focal depth adjustment.

Comparison study of passive acoustic mapping and high-speed photography for monitoring in situ cavitation bubbles.

The spatiotemporal accuracy of passive acoustic mapping (PAM) for monitoring in situ cavitation bubbles has not been assessed directly via optical means. Here, the cavitation bubbles are monitored from two image sequences obtained simultaneously with PAM and high-speed photography (HSP). The temporal accuracy of PAM for detecting cavitation nucleation and the spatial resolution for cavitation localization are compared with those measured from HSP. The results show that PAM has a temporal accuracy of 20 µs. Mean differences in the spatial locations of PAM and HSP are as small as 10.0 and 30.5 µm along the lateral and axial directions, respectively.

Sparse Rectangular and Spiral Array Designs for 3D Medical Ultrasound Imaging.

In three-dimensional (3D) medical ultrasound imaging with two-dimensional (2D) arrays, sparse 2D arrays have been studied to reduce the number of active channels. Among them, sparse 2D arrays with regular or uniform arrangements of elements have advantages of low side lobe energy and uniform field responses over the entire field of view. This paper presents two uniform sparse array models: sparse rectangular arrays (SRAs) on a rectangular grid and sparse spiral arrays (SSAs) on a sunflower grid. Both arrays can be easily implemented on the commercially available or the custom-made arrays. To suppress the overall grating lobe levels, the transmit (Tx) and receive (Rx) array pairs of both the array models are designed not to have grating lobes at the same locations in the Tx/Rx beam patterns, for which the theoretical design rules are also proposed. Computer simulation results indicate that the proposed array pairs for both the SRAs and the SSAs achieve peak grating lobe levels below -40 dB using about a quarter of the number of elements in the dense rectangular array while maintaining similar beam widths to that of the dense array pair.

Monitoring of Adult Zebrafish Heart Regeneration Using High-Frequency Ultrasound Spectral Doppler and Nakagami Imaging.

This paper reports the feasibility of Nakagami imaging in monitoring the regeneration process of zebrafish hearts in a noninvasive manner. In addition, spectral Doppler waveforms that are typically used to access the diastolic function were measured to validate the performance of Nakagami imaging. A 30-MHz high-frequency ultrasound array transducer was used to acquire backscattered echo signal for spectral Doppler and Nakagami imaging. The performances of both methods were validated with flow and tissue-mimicking phantom experiments. For in vivo experiments, both spectral Doppler and Nakagami imaging were simultaneously obtained from adult zebrafish with amputated hearts. Longitudinal measurements were performed for five zebrafish. From the experiments, the E/A ratio measured using spectral Doppler imaging increased at 3 days post-amputation (3 dpa) and then decreased to the value before amputation, which were consistent with previous studies. Similar results were obtained from the Nakagami imaging where the Nakagami parameter value increased at 3 dpa and decreased to its original value. These results suggested that the Nakagami and spectral Doppler imaging would be useful techniques in monitoring the regeneration of heart or tissues.

Ultrasonic sector imaging using plane wave synthetic focusing with a convex array transducer.

Synthetic transmit focusing (STF) methods using unfocused waves or plane waves (PWs) have recently been investigated extensively. However, STF using PWs with a convex array (PWSTF-CA) has not been rigorously studied for high-resolution sector imaging. In this paper, the authors suggest an analytical model for accurate characterization of the spatial beam patterns of PWSTF-CA using a large range of either uniformly or non-uniformly distributed PW angles. On the basis of the model, a frame-based PWSTF-CA approach with non-uniform PW angles is suggested to achieve superior image quality at a higher frame rate than conventional transmit focusing (CTF). The analytical model can also be used for optimal selection of a set of PW angles to scan the entire sectorial field of view and its subsets employed for STF at each imaging point. The authors also investigate how to select transmit subarrays for each of the PWs to obtain the best spatial resolution. A theoretical analysis and simulations are conducted for the verification of the analytical model and the optimal utilization strategy of PWSTF-CA. The results indicate that the PWSTF-CA improves not only the frame rate but also the contrast, signal-to-noise ratio, and resolution compared with the CTF, as in the case of PWSTF with linear arrays.

Real-Time Lossless Compression Algorithm for Ultrasound Data Using BL Universal Code.

Software-based ultrasound imaging systems provide high flexibility that allows easy and fast adoption of newly developed algorithms. However, the extremely high data rate required for data transfer from sensors (e.g., transducers) to the ultrasound imaging systems is a major bottleneck in the software-based architecture, especially in the context of real-time imaging. To overcome this limitation, in this paper, we present a Binary cLuster (BL) code, which yields an improved compression ratio compared to the exponential Golomb code. Owing to the real-time encoding/decoding features without overheads, the universal code is a good solution to reduce the data transfer rate for software-based ultrasound imaging. The performance of the proposed method was evaluated using in vitro and in vivo data sets. It was demonstrated that the BL-beta code has a good stable lossless compression performance of 20%~30% while requiring no auxiliary memory or storage.

Does adding diffuse optical tomography to sonography improve differentiation between benign and malignant breast lesions? Observer performance study.

OBJECTIVES: The purpose of this study was to investigate the added value of diffuse optical tomographic categories combined with conventional sonography for differentiating between benign and malignant breast lesions. METHODS: In this retrospective database review, we included 145 breast lesions (116 benign and 29 malignant) from 145 women (mean age, 46 years; range, 16-86 years). Five radiologists independently reviewed sonograms with and without a diffuse optical tomographic category. Each lesion was scored on a scale of 0% to 100% for suspicion of malignancy and rated according to the American College of Radiology Breast Imaging Reporting and Data System classification. Diagnostic performance was analyzed by comparing area under receiver operating characteristic curve values. Reader agreement was assessed by intraclass correlation coefficients. RESULTS: In the multireader multicase receiver operating characteristic analysis, adding a diffuse optical tomographic category to sonography improved the diagnostic accuracy of sonography (mean areas under the curve, 0.923 for sonography alone and 0.969 for sonography with diffuse optical tomography; P = .039). The interobserver correlation was also improved (0.798 for sonography alone and 0.904 for sonography with diffuse optical tomography). The specificity increased for 4 reviewers from a mean of 19.5% to 45.8% (P < .001 for reviewers 1-4; P = .238 for reviewer 5) with no significant change in the sensitivity. When the diffuse optical tomographic category was applied strictly, the specificity increased for all reviewers from a mean of 19.5% to 68.3% (P < .001 for all reviewers) with no significant change in the sensitivity. CONCLUSIONS: The addition of diffuse optical tomographic categories to sonography may improve diagnostic performance and markedly decrease false-positive biopsy recommendations.

An Adaptive Harmonic Separation Technique for Ultrasound Harmonic Imaging.

OBJECTIVE: An adaptive harmonic separation (HS) technique is proposed to overcome the limitations in conventional filtering techniques for ultrasound (US) tissue harmonic imaging (THI). METHODS: Based on expectation-maximization source separation, the proposed HS technique adaptively models the depth-varying fundamental and harmonic components in the frequency domain and separates the two by applying their calculated posterior probabilities. Phantom experiments with a Tx center frequency of 2 MHz are conducted to evaluate the proposed HS-based US THI schemes. RESULTS: The phantom images show that the proposed single-pulse THI scheme utilizing the HS technique provides not only an average improvement of 19.2% in axial resolution compared to the conventional bandpass filtering scheme but also similar image quality to that of the conventional pulse-inversion (PI) scheme which requires two Tx/Rx sequences for each scan line. Furthermore, when combined with the PI technique, the HS technique provides a uniform axial resolution over the entire 170 mm imaging depth with an average improvement of 17.1% compared to the conventional PI scheme. CONCLUSION: These results show that the proposed adaptive HS technique is capable of improving both the frame rate and the image quality of US THI.

High-contrast spectroscopic photoacoustic characterization of thermal tissue ablation in the visible spectrum.

PURPOSE: High-contrast tissue characterization of thermal ablation has been desired to evaluate therapeutic outcomes accurately. This paper presents a photoacoustic (PA) characterization of thermal tissue ablation in the visible spectrum, in which higher light absorbance can produce spectral contrast starker than in the near-infrared range. METHODS: Ex vivo experiments were performed to measure visible PA spectra (480-700 nm) from fresh porcine liver tissues that received a thermal dose in a range of cumulative equivalent minutes at 43°C (CEM43). The local hemoglobin lobe area between 510-600 nm and wholespectral area under the curve were evaluated to represent the transition of hemoglobin into methemoglobin (MetHb) in the target tissue. RESULTS: The thermal process below an estimated therapeutic CEM43 threshold (80-340 minutes) presented a progressive elevation of the PA spectrum and an eventual loss of local hemoglobin peaks in the visible spectrum, closer to the MetHb spectrum. Interestingly, an excessive CEM43 produced a substantial drop in the PA spectrum. In the spectral analysis, the visible spectrum yielded 13.9-34.1 times higher PA sensitivity and 1.42 times higher contrast change than at a near-infrared wavelength. CONCLUSION: This novel method of PA tissue characterization in the visible spectrum could be a potential modality to evaluate various thermal therapeutic modalities at high-contrast resolution.

Adaptive singular value decomposition filtering of high-intensity focused ultrasound interference enables real-time ultrasound-guided therapy.

High-intensity focused ultrasound (HIFU) is an emerging therapeutic tool for the effective thermal ablation of pathological tissue. For accurate localization of the target and safe control of the HIFU dosage, real-time imaging guidance during the HIFU exposure is desired. Ultrasound imaging has the capability to guide clinicians toward a lesion in real time, but is not an ideal option, as HIFU application causes strong interference, thereby substantially distorting the images used for guidance. Thus, this study introduces singular value decomposition-based filtering capable of restoring ultrasound harmonic images from HIFU interference without undesirable spectral distortion. The results were experimentally validated with a custom-made phantom, indicating that this approach effectively eliminates HIFU-induced artifacts, which is essential for real-time monitoring of the therapeutic process.

Design and Implementation of Analog-Digital Hybrid Beamformers for Low-Complexity Ultrasound Systems: A Feasibility Study.

Low-complexity ultrasound systems are increasingly desired for both wearable, point-of-care ultrasound and high-end massive-channel ultrasound for 3-D matrix imaging. However, the imaging capabilities, including spatial resolution and contrast, could suffer as low complexity systems are pursued, which remains as an unresolved tradeoff. To mitigate this limitation, this study revisits the general structures of analog and digital beamformers and introduces a hybrid approach, referred to as analog-digital hybrid beamforming, to implement efficient ultrasound systems. The suggested hybrid beamforming takes two stages sequentially, where the first analog stage partially beamforms M-channel RF signals to N sum-out data (i.e., M-to-N beamforming), and the second digital stage beamforms N partial sums to single final beamformed data (i.e., N-to-1 beamforming). Our approach was systematically designed and implemented with only four major integrated circuits, which was capable of driving full 64-channel transmission and reception. The developed system was demonstrated with a customized 64-channel 1-D phased array using a commercial tissue mimicking phantom. From the phantom imaging results, signal-to-noise ratio, contrast-to-noise ratio, and full beam width at half maximum values were quantitatively evaluated. The demonstrated results indicate that the analog-digital hybrid beamforming can be applied to any type of array for sophisticated 3-D imaging and tiny wearable ultrasound applications.

Development of exosome membrane materials-fused microbubbles for enhanced stability and efficient drug delivery of ultrasound contrast agent.

Lipid-coated microbubbles are widely used as an ultrasound contrast agent, as well as drug delivery carriers. However, the two main limitations in ultrasound diagnosis and drug delivery using microbubbles are the short half-life in the blood system, and the difficulty of surface modification of microbubbles for active targeting. The exosome, a type of extracellular vesicle, has a preferentially targeting ability for its original cell. In this study, exosome-fused microbubbles (Exo-MBs) were developed by embedding the exosome membrane proteins into microbubbles. As a result, the stability of Exo-MBs is improved over the conventional microbubbles. On the same principle that under the exposure of ultrasound, microbubbles are cavitated and self-assembled into nano-sized particles, and Exo-MBs are self-assembled into exosome membrane proteins-embedded nanoparticles (Exo-NPs). The Exo-NPs showed favorable targeting properties to their original cells. A photosensitizer, chlorin e6, was loaded into Exo-MBs to evaluate therapeutic efficacy as a drug carrier. Much higher therapeutic efficacy of photodynamic therapy was confirmed, followed by cancer immunotherapy from immunogenic cell death. We have therefore developed a novel ultrasound image-guided drug delivery platform that overcomes the shortcomings of the conventional ultrasound contrast agent and is capable of simultaneous photodynamic therapy and cancer immunotherapy.

Enhancement of photoacoustic image quality by sound speed correction: ex vivo evaluation.

Real-time photoacoustic (PA) imaging involves beamforming methods using an assumed fixed sound speed, typically 1540 m/s in soft tissue. This leads to degradation of PA image quality because the true sound speed changes as PA signal propagates through different types of soft tissues: the range from 1450 m/s to 1600 m/s. This paper proposes a new method for estimating an optimal sound speed to enhance the cross-sectional PA image quality. The optimal sound speed is determined when coherent factor with the sound speed is maximized. The proposed method was validated through simulation and ex vivo experiments with microcalcification-contained breast cancer specimen. The experimental results demonstrated that the best lateral resolution of PA images of microcalcifications can be achieved when the optimal sound speed is utilized.

Elevational Synthetic Aperture Focusing for Three-Dimensional Photoacoustic Imaging Using a Clinical One-Dimensional Array Transducer.

OBJECTIVE: Two-dimensional (2D) photoacoustic (PA) imaging based on array transducers provide high spatial resolution in the lateral direction by adopting receive dynamic focusing. However, the quality of PA image is often deteriorated by poor elevational resolution which is achieved by an acoustic lens. To overcome this limitation, we present a three-dimensional (3D) image reconstruction method using a commercial one-dimensional (1D) array transducer. METHODS: In the method, the elevational resolution is improved by applying synthetic aperture focusing (SAF) technique along the elevational direction. For this, a commercially available 1D array transducer with an acoustic lens is modeled and appropriate synthetic focusing delay that can minimize the effect of the acoustic lens is derived by mathematical analysis. RESULTS: From the simulation and experiment results, it was demonstrated that the proposed method can enhance the image quality of PA imaging, i.e., elevational resolution and signal-to-noise ratio (SNR). CONCLUSION: 3D PA images with improved elevational resolution were achieved using a clinical 1D array transducer. SIGNIFICANCE: The presented method may be useful for clinical application such as detecting microcalcification, imaging of tumor vasculature and guidance of biopsy in real time.

Experimental Study of Aperiodic Plane Wave Imaging for Ultrafast 3-D Ultrasound Imaging.

OBJECTIVE: Although plane wave imaging (PWI) with multiple plane waves (PWs) steered at different angles enables ultrafast three-dimensional (3-D) ultrasonic imaging, there is still a challenging tradeoff between image quality and frame rate. To address this challenge, we recently proposed the aperiodic PWI (APWI) with mathematical analysis and simulation study. In this paper, we demonstrate the feasibility of APWI and evaluate the performance with phantom and in vivo experiments. METHODS: APWI with a concentric ring angle pattern (APWI-C) and APWI with a sunflower pattern (APWI-S) are evaluated. For experimental verification of the methods, the experimental results are compared with simulation results in terms of the spatial resolution and the mainlobe-to-sidelobe ratio. In addition, the performance of APWI is compared with that of conventional PWI by using a commercial phantom. To examine the potential for clinical use of APWI, a gallstone phantom study and an in vivo carotid artery experiment are also conducted. RESULTS: In the phantom study, the APWI methods provide a contrast ratio approximately 2-3 dB higher than that of PWI. In a gallstone experiment, the proposed methods yield 3-D rendered stone images more similar to the real stones than PWI. In the in vivo carotid artery images, APWI reduces the clutter artifacts inside the artery. CONCLUSION: Phantom and in vivo studies show that the APWI enhances the contrast without compromising the spatial resolution and frame rate. SIGNIFICANCE: This study experimentally demonstrates the feasibility and advantage of APWI for ultrafast 3-D ultrasonic imaging.

Precursor Heterogeneity Driven Mo-Te Nanoparticle Structural Diversification for Cancer Photo-Theranostics.

Chemical reactions between homogeneous precursors are typically used to synthesize monodisperse nanoparticles with well-controlled size and morphology. It is difficult to predict the evolved nanostructures when using two heterogeneous precursors. In this study, three types of Mo-Te nanoparticles shaped like leaves, spindles, and rice grains (denoted respectively as nanoleaf, nanospindle, and nanorice) were obtained from dextrose-mediated proton-coupled electron transfer reaction between the solid polyoxomolybdate (POM) and the ionic tellurite anion as precursors. All produced nanoparticles had excellent optical absorption in the ultraviolet(UV)-visible(Vis)-near-infrared(NIR) regions, with only slight deviations among them. After confirming nanoparticles' photothermal conversion and photocatalytic activity at multiple wavelengths, the Mo-Te nanorice was tested as a potential agent for cancer treatment due to its minimum toxicity, excellent colloidal stability, and intrinsic anticancer effect. Excellent treatment efficacy and clearance were confirmed in vitro and in vivo. Due to their photoacoustic imaging capability, the injection of pristine nanoparticles could also realize phototheranostics without using additional drugs, probes, or photosensitizers.

Analysis of the Time and Phase Delay Resolutions in Ultrasound Baseband I/Q Beamformers.

OBJECTIVE: The ultrasound baseband in-phase/quadrature beamformer (IQBF) has been widely employed in medical ultrasound imaging to reduce the amount of channel data or to decrease the data rate of the beamforming process. The aim of this study is to assess the effect of the time and phase delay compensation accuracies on the IQBF and thereby to suggest the criteria for selecting the delay resolutions of the IQBF. METHODS: Mathematical models of the gain loss (GL) and sidelobe level (SL) in closed form are suggested, and the relationships between the parameters (time and phase delay resolutions of the IQBF and the signal bandwidth) and the errors (GL and SL) are investigated. The performance of the IQBF is compared with that of the traditional radio-frequency beamformer (RFBF). Simulation and phantom and in vivo experimental results are shown to corroborate the theoretical analysis. RESULTS AND CONCLUSION: Theoretical analysis and simulation and experimental results show that a phase delay resolution with a quantization step of 2π/64 is sufficient for phase compensation and that a time delay resolution with a sampling rate of 4f0 and 2f0 in the IQBF is sufficient for data with a -6 dB bandwidth of 50% and 25%, respectively, for similar performance as the RFBF with a sampling rate of 16f0, where f0 is the center frequency of the ultrasound signal. SIGNIFICANCE: The suggested criteria have the potential to be used for designing an efficient IQBF satisfying the desired specifications and beamforming accuracy.

A new frequency domain passive acoustic mapping method using passive Hilbert beamforming to reduce the computational complexity of fast Fourier transform.

Passive acoustic mapping (PAM) is the current state-of-the-art imaging tool for monitoring cavitation activity during focused ultrasound therapy such as blood-brain barrier opening. However, PAM incurs huge computational complexity. To address this issue, frequency-domain PAM (FD-PAM) was proposed. Nevertheless, FD-PAM still requires a large number of fast Fourier transforms (FFTs) to produce the frequency components utilized for cavitation monitoring with PAM. Hence, in this paper, we proposes a frequency domain PAM method using passive Hilbert beamforming (PHB-PAM), which can significantly reduce the number of input samples for FFT by down-sampling the analytic signal of the received RF samples at each channel at a rate equal to the bandwidth of the frequency components of interest. The experimental results show that the proposed PHB-PAM provides comparable image quality to that of FD-PAM (correlation coefficient > 0.98). Additionally, the study experimentally verifies that the pre-processing block for generating the decimated analytic signal and FFT in PHB-PAM can be realized using lesser logic resources than FFT in FD-PAM when implemented in an FPGA. Especially, with 128-fold decimation, PHB-PAM reduces the amount of LUTs and DSP slices to implement the pre-processing block by 72.16% and 53.4%, respectively, compared to those of FD-PAM, which allows the 64-channel implementation of the pre-processing block in a low-cost single FPGA. Finally, a hardware-efficient architecture for the pre-processing block of PHB-PAM is described, which can be implemented by replacing the two lowpass filters of an off-the-shelf analog front-end component for ultrasound imaging with a pair of band-pass filters. If PHB-PAM is realized using such a component, it can truly minimize the computational complexity of FD-PAM.