Raw Radiofrequency Research Articles

Radio Frequency Signal-Based Drone Classification with Frequency Domain Gramian Angular Field and Convolutional Neural Network

Over the past few years, drones have been utilized in a wide range of applications. However, the illegal operation of drones may pose a series of security risks to sensitive areas such as airports and military bases. Hence, it is vital to develop an effective method of identifying drones to address the above issues. Existing drone classification methods based on radio frequency (RF) signals have low accuracy or a high computational cost. In this paper, we propose a novel RF signal image representation scheme that incorporates a convolutional neural network (CNN), named the frequency domain Gramian Angular Field with a CNN (FDGAF-CNN), to perform drone classification. Specifically, we first compute the time–frequency spectrum of raw RF signals based on short-time Fourier transform (STFT). Then, the 1D frequency spectrum series is encoded as 2D images using a modified GAF transform. Moreover, to further improve the recognition performance, the images obtained from different channels are fused to serve as the input of a CNN classifier. Finally, numerous experiments were conducted on the two available open-source DroneRF and DroneRFa datasets. The experimental results show that the proposed FDGAF-CNN can achieve a relatively high classification accuracy of 98.72% and 98.67% on the above two datasets, respectively, confirming the effectiveness and generalization ability of the proposed method.

Aberration correction by polynomial approximation for synthetic aperture ultrasound imaging

AbstractBackgroundPreviously, there has been some work in the field of optical imaging on phase compensation by employing Legendre polynomials as an expansion of the phase function, and this seems to be an appealing unstudied area in the field of ultrasound imaging.PurposeThe paper is devoted to solving one of the problems of enhancing the authenticity of diagnostics data obtained in ultrasound visualization systems and presents a novel approach to aberration correction, which ensures a reliable eliminating of phase distortions. The novelty of the proposed approach consists in the use of the decomposition of the wave front by Legendre polynomials to approximate aberrations of the phase front of ultrasonic waves propagating through distorting layers.MethodsThe phase aberrations are corrected using a single sector probe that captures echoes in synthetic aperture mode. The proposed method approximates the changes in the wave front by means of the Legendre polynomials. The performance was investigated by placing a 13‐mm‐wide ultrasonic probe with 64 elements operating at 2 MHz on the surface of a commercial quality control phantom ATS Model 539 through a distorting layer. The following metrics were measured: peak value, root mean square (RMS), full width at half maximum (FWHM), contrast‐to‐noise ratio (CNR). During the experiments, three different aberrators were used interchangeably as distorting layers, two of which were made of photopolymer resin with RMS values of 39 and 97 ns and a speed of sound close to that of a human temporal bone tissue, and the third was an ex vivo human temporal bone with the RMS value of 44 ns. To test the correction, three regions inside the phantom were examined. Two‐sided nonparametric Wilcoxon rank‐sum test was used for assessing the statistical significance. The significance level for this study was set at α ≤ 0.05. To counteract the problem of multiple comparisons, the p‐values were adjusted by the Holm–Bonferroni correction method.ResultsThe statistically significant results have demonstrated the possibility of increasing peak intensity value up to 3.08 times, reducing the RMS width and FWHM of the intensity angular distribution down to 60% and 82%, respectively, and increasing CNR up to 2.12 times by using the proposed phase distortion correction method, compared with the case without correction. The results show that the method can be used as an effective aberration correction technique for all tested regions inside the phantom and distorting layers. Its usage allowed correcting up to 97% of the aberrations caused by the ex vivo temporal bone model.ConclusionThe results show that the usage of the proposed method for aberration correction can successfully increase the intensity and reduce the angular width of the ultrasound wave scattered by the point targets inside the phantoms. The method works with different distorting layers and is capable of correcting phase aberrations in multiple sections of the sonogram. The limitation of the proposed method consists in the fact that it requires the use of aperture synthesis and access to raw radiofrequency data, which restricts its application in common scanners.

Enhanced identification of women at risk for preterm birth via quantitative ultrasound: a prospective cohort study

Historically, clinicians have relied on medical risk factors and clinical symptoms for preterm birth risk assessment. In nulliparous women, clinicians may rely solely on reported symptoms to assess for the risk of preterm birth. The routine use of ultrasound during pregnancy offers the opportunity to incorporate quantitative ultrasound scanning of the cervix to potentially improve assessment of preterm birth risk. This study aimed to investigate the efficiency of quantitative ultrasound measurements at relatively early stages of pregnancy to enhance identification of women who might be at risk for spontaneous preterm birth. A prospective cohort study of pregnant women was conducted with volunteer participants receiving care from the University of Illinois Hospital in Chicago, Illinois. Participants received a standard clinical screening followed by 2 research screenings conducted at 20±2 and 24±2 weeks. Quantitative ultrasound scans were performed during research screenings by registered diagnostic medical sonographers using a standard cervical length approach. Quantitative ultrasound features were computed from calibrated raw radiofrequency backscattered signals. Full-term birth outcomes and spontaneous preterm birth outcomes were included in the analysis. Medically indicated preterm births were excluded from the analysis. Using data from each visit, logistic regression with Akaike information criterion feature selection was conducted to derive predictive models for each time frame based on historical clinical and quantitative ultrasound features. Model evaluations included a likelihood ratio test of quantitative ultrasound features, cross-validated receiver operating characteristic curve analysis, sensitivity, and specificity. On the basis of historical clinical features alone, the best predictive model had an estimated receiver operating characteristic area under the curve of 0.56±0.03. By the time frame of Visit 1, a predictive model using both historical clinical and quantitative ultrasound features provided a modest improvement in the area under the curve (0.63±0.03) relative to that of the predictive model using only historical clinical features. By the time frame of Visit 2, the predictive model using historical clinical and quantitative ultrasound features provided significant improvement (likelihood ratio test, P<.01), with an area under the curve of 0.69±0.03. Accurate identification of women at risk for spontaneous preterm birth solely through historical clinical features has been proven to be difficult. In this study, a history of preterm birth was the most significant historical clinical predictor of preterm birth risk, but the historical clinical predictive model performance was not statistically significantly better than the no-skill level. According to our study results, including quantitative ultrasound yields a statistically significant improvement in risk prediction as the pregnancy progresses.

Ultrasound tissue scatterer distribution imaging: An adjunctive diagnostic tool for shear wave elastography in characterizing focal liver lesions

ObjectivesFocal liver lesion (FLL) is a prevalent finding in cross-sectional imaging, and distinguishing between benign and malignant FLLs is crucial for liver health management. While shear wave elastography (SWE) serves as a conventional quantitative ultrasound tool for evaluating FLLs, ultrasound tissue scatterer distribution imaging (TSI) emerges as a novel technique, employing the Nakagami statistical distribution parameter to estimate backscattered statistics for tissue characterization. In this prospective study, we explored the potential of TSI in characterizing FLLs and evaluated its diagnostic efficacy with that of SWE. MethodsA total of 235 participants (265 FLLs; the study group) were enrolled to undergo abdominal examinations, which included data acquisition from B-mode, SWE, and raw radiofrequency data for TSI construction. The area under the receiver operating characteristic curve (AUROC) was used to evaluate performance. A dataset of 20 patients (20 FLLs; the validation group) was additionally acquired to further evaluate the efficacy of the TSI cutoff value in FLL characterization. ResultsIn the study group, our findings revealed that while SWE achieved a success rate of 49.43 % in FLL measurements, TSI boasted a success rate of 100 %. In cases where SWE was effectively implemented, the AUROCs for characterizing FLLs using SWE and TSI stood at 0.84 and 0.83, respectively. For instances where SWE imaging failed, TSI achieved an AUROC of 0.78. Considering all cases, TSI presented an overall AUROC of 0.81. There was no statistically significant difference in AUROC values between TSI and SWE (p > 0.05). In the validation group, using a TSI cutoff value of 0.67, the AUROC for characterizing FLLs was 0.80. ConclusionsIn conclusion, ultrasound TSI holds promise as a supplementary diagnostic tool to SWE for characterizing FLLs.

Spectral-based quantitative ultrasound for assessing vision degrading myodesopsia

The presence of clinically significant vitreous opacities, referred to as vision degrading myodesopsia (VDM), reduces contrast sensitivity (CS), and visual quality of life. Quantitative ultrasound (QUS) methods operating on grayscale B-mode images were previously developed and validated for quantifying vitreous echodensities. Although a subsequent study comparing the QUS methods across clinical scanners demonstrated no change in QUS efficacy, the methods are sensitive to differences in system parameters and B-mode image formation unless the backscatter data is normalized. Technological advancements in clinical scanners allow for easy access to raw radiofrequency (RF) data for more advanced processing methods. Here, we develop QUS methods that exploit RF echo data to generate robust user- and system-independent parameters for quantifying vitreous echodensities based on the echo signal power spectrum (PS). However, the sparsity of echodensities in the vitreous body is incompatible with underlying assumptions of QUS methods based on measurements of the backscatter coefficient or envelope statistics. We, therefore, implement QUS methods that parameterize the PS and correlate these parameters with visual function as measured by CS. Results from this study are the first step toward developing and validating robust, user-, and system-independent QUS methods for evaluating VDM.

Initial Study of A-mode Ultrasound Spectroscopy Through Mechanical Wave Scattering Phenomenon for Measuring 3D-printed Bone Model Density

In Indonesia, the prevalence of osteoporosis is high. Given the economic burden it may impose on the population, this condition must be taken seriously. Dual-energy X-ray absorptiometry is the gold standard for diagnosing osteoporosis (DEXA). However, due to its high cost, non-portability, and radiation risk, DEXA cannot be applied to large populations. An alternative method for evaluating bone quality is ultrasound. It is more affordable, portable, and has no radiation risk. In this preliminary study, an A-mode ultrasound spectroscopy prototype for assessing the density of a 3D-printed bone model is designed. A single-element transducer (Transmit-Tx/Receive-Rx), a reconfigurable and modular FPGA-based ultrasound beamformer system, and a Raspberry Pi 3 are the system's control units. The raw radio frequency (RF) signal is acquired from three variations of density of the 3D-printed bone model, i.e., 100%, 60%, and 40%, to represent normal bone, osteopenia, and osteoporosis. The designed prototype can adequately characterize the mechanical wave scattering pattern of the 3D-printed bone model indicated by the increased tendency in the maximum amplitude when the density of the bone model is increasing. The tendency is the opposite for delay time and Power Spectral Density (PSD). These three signal parameters are potential candidate parameters to represent bone density. For future work, the selected candidate parameters can later be used as reference values while adding a significant data so that a machine learning method can be employed to extract representative features of bone density level, i.e., normal bone, osteopenia, and osteoporosis.

High-Level Synthesis Design of Scalable Ultrafast Ultrasound Beamformer With Single FPGA.

Ultrafast ultrasound imaging is essential for advanced ultrasound imaging techniques such as ultrasound localization microscopy (ULM) and functional ultrasound (fUS). Current ultrafast ultrasound imaging is challenged by the ultrahigh data bandwidth associated with the radio frequency (RF) signal, and by the latency of the computationally expensive beamforming process. As such, continuous ultrafast data acquisition and beamforming remain elusive with existing software beamformers based on CPUs or GPUs. To address these challenges, the proposed work introduces a novel method of implementing an ultrafast ultrasound beamformer specifically for ultrafast plane wave imaging (PWI) on a field programmable gate array (FPGA) by using high-level synthesis. A parallelized implementation of the beamformer on a single FPGA was proposed by 1) utilizing a delay compression technique to reduce the delay profile size, which enables both run-time pre-calculated delay profile loading from external memory and delay reuse, 2) vectorizing channel data fetching which is enabled by delay reuse, and 3) using fixed summing networks to reduce consumption of logic resources. Our proposed method presents two unique advantages over current FPGA beamformers: 1) high scalability that allows fast adaptation to different FPGA resources and beamforming speed demands by using Xilinx High-Level Synthesis as the development tool, and 2) allow a compact form factor design by using a single FPGA to complete the beamforming instead of multiple FPGAs. Current Xilinx FPGAs provide the capabilities of connecting up to 1024 ultrasound channels with a single FPGA and the newest JESD204B interface analog front end (AFE). This channel count is much more than the channel count needed by current linear arrays, which normally have 128 or 256 channels. With the proposed method, a sustainable average beamforming rate of 4.83 G samples/second in terms of input raw RF sample was achieved. The resulting image quality of the proposed beamformer was compared with the software beamformer on the Verasonics Vantage system for both phantom imaging and in vivo imaging of a mouse brain. Multiple imaging schemes including B-mode, power Doppler and ULM were assessed to verify that the image quality was not compromised for speed.

Skeletal Muscle Assessment Using Quantitative Ultrasound: A Narrative Review.

Ultrasound (US) is an important imaging tool for skeletal muscle analysis. The advantages of US include point-of-care access, real-time imaging, cost-effectiveness, and absence of ionizing radiation. However, US can be highly dependent on the operator and/or US system, and a portion of the potentially useful information carried by raw sonographic data is discarded in image formation for routine qualitative US. Quantitative ultrasound (QUS) methods provide analysis of the raw or post-processed data, revealing additional information about normal tissue structure and disease status. There are four QUS categories that can be used on muscle and are important to review. First, quantitative data derived from B-mode images can help determine the macrostructural anatomy and microstructural morphology of muscle tissues. Second, US elastography can provide information about muscle elasticity or stiffness through strain elastography or shear wave elastography (SWE). Strain elastography measures the induced tissue strain caused either by internal or external compression by tracking tissue displacement with detectable speckle in B-mode images of the examined tissue. SWE measures the speed of induced shear waves traveling through the tissue to estimate the tissue elasticity. These shear waves may be produced using external mechanical vibrations or internal "push pulse" ultrasound stimuli. Third, raw radiofrequency signal analyses provide estimates of fundamental tissue parameters, such as the speed of sound, attenuation coefficient, and backscatter coefficient, which correspond to information about muscle tissue microstructure and composition. Lastly, envelope statistical analyses apply various probability distributions to estimate the number density of scatterers and quantify coherent to incoherent signals, thus providing information about microstructural properties of muscle tissue. This review will examine these QUS techniques, published results on QUS evaluation of skeletal muscles, and the strengths and limitations of QUS in skeletal muscle analysis.

Boosting multi‐target recognition performance with multi‐input multi‐output radar‐based angular subspace projection and multi‐view deep neural network

AbstractCurrent radio frequency (RF) classification techniques assume only one target in the field of view. Multi‐target recognition is challenging because conventional radar signal processing results in the superposition of target micro‐Doppler signatures, making it difficult to recognise multi‐target activity. This study proposes an angular subspace projection technique that generates multiple radar data cubes (RDC) conditioned on angle (RDC‐ω). This approach enables signal separation in the raw RDC, making possible the utilisation of deep neural networks taking the raw RF data as input or any other data representation in multi‐target scenarios. When targets are in closer proximity and cannot be separated by classical techniques, the proposed approach boosts the relative signal‐to‐noise ratio between targets, resulting in multi‐view spectrograms that boosts the classification accuracy when input to the proposed multi‐view DNN. Our results qualitatively and quantitatively characterise the similarity of multi‐view signatures to those acquired in a single‐target configuration. For a nine‐class activity recognition problem, 97.8% accuracy in a 3‐person scenario is achieved, while utilising DNN trained on single‐target data. We also present the results for two cases of close proximity (sign language recognition and side‐by‐side activities), where the proposed approach has boosted the performance.

Predicting Spontaneous Pre-term Birth Risk Is Improved When Quantitative Ultrasound Data Are Included With Historical Clinical Data.

Predicting women at risk for spontaneous pre-term birth (sPTB) has been medically challenging because of the lack of signs and symptoms of pre-term birth until interventions are too late. We hypothesized that prediction of the sPTB risk level is enhanced when using both historical clinical (HC) data and quantitative ultrasound (QUS) data compared with using only HC data. HC data defined herein included birth history prior to that of the current pregnancy as well as, from the current pregnancy, a clinical cervical length assessment and physical examination data. The study population included 248 full-term births (FTBs) and 26 sPTBs. QUS scans (Siemens S2000 and MC9-4) were performed by registered diagnostic medical sonographers using a standard cervical length approach. Two cervical QUS scans were conducted at 20 ± 2 and 24 ± 2 wk of gestation. Multiple QUS features were evaluated from calibrated raw radiofrequency backscattered ultrasonic signals. Two statistical models designed to determine sPTB risk were compared: (i) HC data alone and (ii) combined HC and QUS data. Model comparisons included a likelihood ratio test, cross-validated receiver operating characteristic area under the curve, sensitivity and specificity. The study's birth outcomes were only FTBs and sPTBs; medically induced pre-term births were not included. Combined HC and QUS data identified women at risk of sPTB with better AUC (0.68, 95% confidence interval [CI]: 0.57-0.78) compared with HC data alone (0.53, 95% CI: 0.40-0.66) and HC data+cervical length at 18-20 wk of gestation (average AUC=0.51, 95% CI: 0.38-0.64). A likelihood ratio test for significance of QUS features in the classification model was highly statistically significant (p < 0.01). Even with only 26 sPTBs among 274 births, value was added in predicting sPTB when QUS data were included with HC data.

Evaluating the feasibility of differentiating glioblastoma and normal brain tissue using H-scan imaging: An ex vivo study

H-scan imaging is a quantitative ultrasound technique in which raw radiofrequency (RF) echoes are matched to Gaussian-weighted Hermite polynomials of order n, which is related to the size of scatterers. This information is used to characterize tissue structures based on color-mapping large scatterers in the red channel and small scatterers in the blue channel of the H-scan image. Glioblastoma is a malignant tumor of the brain and has one of the most dismal prognoses of all tumors. In this study, we assessed the feasibility of differentiating glioblastoma and normal brain tissue ex vivo using H-scan imaging. Raw RF echoes from ex vivo brain tissues embedded in agarose phantoms were acquired using a Verasonics Vantage 128 system equipped with an L11-5v array. A total of 11 samples each of glioblastoma and normal tissues were evaluated. The intensity-weighted percentage of thered channel (IWPRED), a scatterer size-dependent and concentration-independentparameter, was computed for each sample. The IWPRED parameter estimates of glioblastoma and normal tissue samples were significantly different (24.0% ± 0.7% vs 34.2% ± 1.1%, respectively; p &amp;lt; 0.001). This work demonstrates the feasibility of differentiating glioblastoma tumors from normal tissue using H-scan imaging.

Quantitative lung ultrasound spectroscopy, an in vivo clinical study conducted over 101 patients

Lung ultrasound (LUS) is widely adopted to assess the state of the lung surface. However, as standard ultrasound imaging assumptions are unmet in the lung due to the presence of air, essentially this organ cannot be anatomically investigated. Indeed, LUS is mainly based on the analysis of imaging artifacts (horizontal and vertical). Of particular interest is the analysis of vertical artifacts, as they correlate with several pathologies. More specifically, their dependence on frequency was demonstrated to carry important diagnostic information capable to improve LUS specificity. In this study, a new dataset was generated acquiring raw radiofrequency (RF) data from 101 patients (affected by different pathologies, e.g., COVID-19 pneumonia and cardiogenic edema). The ULA-OP research platform was used to collect the data, and a multifrequency approach implemented on both linear and convex probes was adopted. Ultrasound data were thus acquired utilizing orthogonal sub-bands with a 1-MHz bandwidth and with different center frequencies (2–6 MHz). About 12 000 and 3600 multifrequency frames were acquired with convex and linear probes, respectively. Clinical parameters (e.g., FIO2 and LDH) were also stored for all patients. During this meeting, the dataset will be presented together with preliminary results, and its potential applications discussed.

Resolution and contrast improved ultrafast power Doppler microvessel imaging with null subtraction imaging

In this work, we demonstrate a novel non-linear beamforming technique, called null-subtraction imaging (NSI), to improve spatial resolution and image contrast of ultrafast, power Doppler microvessel imaging without significantly increasing computational cost compared to traditional delay and sum (DAS). NSI is a non-linear beamforming technique that operates on receive data by subtracting envelopes of beamforming results from three apodization windows. An array transducer and Verasonics Vantage 256 system were used to capture 1600 frames of raw ultrasonic radio frequency (RF) channel data by scanning a rat brain using nine plane waves spanning angles between -4° and 4° at a pulse repetition frequency of 1000 Hz. A singular value decomposition filter was applied on the raw RF channel data to filter out the tissue signal. Coherent compounding was performed individually on the beamformed result for each apodization window. Next, the subtraction of the envelopes for each apodization was conducted to form the NSI image. A total of 1,600 frames of NSI images were accumulated to form the final image. The NSI based ultrafast power Doppler microvessel imaging provided a ten-fold improvement in spatial resolution and a three-fold improvement in image contrast compared with the traditional DAS-based ultrafast power Doppler microvessel imaging.

Predicting spontaneous PTB risk is improved when quantitative ultrasound data are included with clinical data

Predicting women at risk for spontaneous preterm birth (sPTB) has been medically challenging due to the lack of signs and symptoms of preterm labor until intervention is too late. Hypothesis: Predicting the sPTB risk level is enhanced when using both historical clinical (HC) data and quantitative ultrasound (QUS) data compared to using only HC data, with HC data being defined as birth history prior current pregnancy as well as, from the current pregnancy, a clinical cervical length assessment, and data from a physical exam. Study population included 250 full-term births (FTBs) and 25 sPTBs. QUS scans (Siemens S2000 & MC9-4) were performed using a standard cervical length approach by RDMS sonographers. QUS is a measure of cervical remodeling at the microstructure level. Two cervical QUS scans were conducted at 20±2 and 24±2 weeks’ gestation. Multiple QUS features were evaluated from calibrated raw radiofrequency backscattered ultrasonic signals. Two statistical models designed to determine sPTB risk were compared: 1) HC data alone and 2) combined HC and QUS data. Test ROC AUC compared both models. The study’s birth outcomes were only FTBs or sPTBs; medically induced preterm births were not included. Combined HC and QUS data identified women at sPTB risk with better AUC (0.68; 95% CI, 0.58-0.78) than that of HC data alone (0.52; 95% CI, 0.38-0.65). A likelihood ratio test for significance of QUS features in the classification model was highly statistically significant (p < 0.01). Traditional risk factors for sPTB (age, parity, previous history of PTB, cervical length < 25 mm, smoking, BMI and alcohol use) were not predictive of sPTB in our sample of participants. Cervical length measurements and HC have been the mainstay of risk assessment for sPTB for years because no other credible biomarkers were available. Even with only 25 sPTBs, there was value-added for predicting sPTB when QUS data were included with HC data.

In Vivo Longitudinal Monitoring of Cardiac Remodeling in Murine Ischemia Models With Adaptive Bayesian Regularized Cardiac Strain Imaging: Validation Against Histology.

Adaptive Bayesian regularized cardiac strain imaging (ABR-CSI) uses raw radiofrequency signals to estimate myocardial wall contractility as a surrogate measure of relative tissue elasticity incorporating regularization in the Bayesian sense. We determined the feasibility of using ABR-CSI -derived strain for in vivo longitudinal monitoring of cardiac remodeling in a murine ischemic injury model (myocardial infarction [MI] and ischemia-reperfusion [IR]) and validated the findings against ground truth histology. We randomly stratified 30 BALB/CJ mice (17 females, 13 males, median age=10 wk) into three surgical groups (MI=10, IR=12, sham=8) and imaged pre-surgery (baseline) and 1, 2, 7 and 14 d post-surgery using a pre-clinical high-frequency ultrasound system (VisualSonics Vevo 2100). We then used ABR-CSI to estimate end-systolic and peak radial (er) and longitudinal (el) strain estimates. ABR-CSI was found to have the ability to serially monitor non-uniform cardiac remodeling associated with murine MI and IR non-invasively through temporal variation of strain estimates post-surgery. Furthermore, radial end-systole (ES) strain images and segmental strain curves exhibited improved discrimination among infarct, border and remote regions around the myocardium compared with longitudinal strain results. For example, the MI group had significantly lower (Friedman's with Bonferroni-Dunn test, p=0.002) ES er values in the anterior middle (infarcted) region at day 14 (n=9, 9.23 ± 7.39%) compared with the BL group (n=9, 44.32 ± 5.49). In contrast, anterior basal (remote region) mean ES er values did not differ significantly (non-significant Friedman's test, χ2=8.93, p=0.06) at day 14 (n=6, 33.05 ± 6.99%) compared with baseline (n=6, 34.02 ± 6.75%). Histology slides stained with Masson's trichrome (MT) together with a machine learning model (random forest classifier) were used to derive the ground truth cardiac fibrosis parameter termed histology percentage of myocardial fibrosis (PMF). Both radial and longitudinal strain were found to have strong statistically significant correlations with the PMF parameter. However, radial strain had a higher Spearman's correlation value (εresρ=-0.67, n=172, p < 0.001) compared with longitudinal strain (εlesρ=-0.60, n=172, p < 0.001). Overall, the results of this study indicate that ABR-CSI can reliably perform non-invasive detection of infarcted and remote myocardium in small animal studies.

Evaluation of Takayasu's arteritis activity by ultrasound localization microscopy

Abstract Objectives Use of sulphur hexafluoride microbubbles (MB) with ultrasound can provide arterial wall enhancement. The use of ultrafast imaging with MB super-localization offers the possibility of ultrasound localization microscopy (ULM) and thus a visualization of in vivo carotid vasa vasorum. We aim to perform in vivo Vasa vasorum imaging in the carotid wall in Takayasu's arteritis and to provide a correlation with the disease's activity. Methods Patients with Takayasu arteritis were consecutively included in the national referral centre. Assessment of activity was performed by NIH activity index, with measurement of biological inflammation, morphological evaluation by computed tomography (CT) angiography, and 18-fluorodeoxyglucose CT scan. Ultrafast ultrasound imaging was performed with a 7 MHz central frequency linear probe. A dedicated contrast imaging sequence (plane waves with 8 angles, frame rate 500 Hz, voltage 8 V, duration of 8 s), coupled with the intravenous injection of sulphur hexafluoride MB allow us to store the raw radio frequency data of the MB passage. After a pre-processing step to accumulate frames properly including a frames selection and a movement correction, the MB were localized then tracked, allowing reconstruction of the vasa vasorum. The inflammation was quantified through MB tracked and normalized for inter acquisition comparison. Results 16 patients were included. 5 patients (median age 35.8 [24.5–46.0] years, 3 women) had active disease and 11 patients (37.2 [31.7–47.3] years, 9 women) had quiescent disease. The passage of MB allowed the visualisation of microvessels within the arterial wall for active cases. The number of MB detected per second in the wall was 118 [80–169] for active cases vs. 13 [10–15] for quiescent cases (p=0.0005). Conclusion ULM allows visualization of microvessels within the carotid wall, with significantly greater MB passage in active Takayasu arteritis. ULM provides for the first time a precise visualization in vivo of the vasa vasorum and gives access to quantification of the vascularity of the arterial wall. Funding Acknowledgement Type of funding sources: Foundation. Main funding source(s): Société française de cardiologie

Prediction of response to neoadjuvant chemotherapy in breast cancer with recurrent neural networks and raw ultrasound signals

Objective. Prediction of the response to neoadjuvant chemotherapy (NAC) in breast cancer is important for patient outcomes. In this work, we propose a deep learning based approach to NAC response prediction in ultrasound (US) imaging. Approach. We develop recurrent neural networks that can process serial US imaging data to predict chemotherapy outcomes. We present models that can process either raw radio-frequency (RF) US data or regular US images. The proposed approach is evaluated based on 204 sequences of US data from 51 breast cancers. Each sequence included US data collected before the chemotherapy and after each subsequent dose, up to the 4th course. We investigate three pre-trained convolutional neural networks (CNNs) as back-bone feature extractors for the recurrent network. The CNNs were pre-trained using raw US RF data, US b-mode images and RGB images from the ImageNet dataset. The first two networks were developed using US data collected from malignant and benign breast masses. Main results. For the pre-treatment data, the better performing network, with back-bone CNN pre-trained on US images, achieved area under the receiver operating curve (AUC) of 0.81 (±0.04). Performance of the recurrent networks improved with each course of the chemotherapy. For the 4th course, the better performing model, based on the CNN pre-trained with RGB images, achieved AUC value of 0.93 (±0.03). Statistical analysis based on the DeLong test presented that there were no significant differences in AUC values between the pre-trained networks at each stage of the chemotherapy (p-values > 0.05). Significance. Our study demonstrates the feasibility of using recurrent neural networks for the NAC response prediction in breast cancer US.

Contrast-enhanced ultrasound tracking of helical propellers with acoustic phase analysis and comparison with color Doppler.

Medical microrobots (MRs) hold the potential to radically transform several interventional procedures. However, to guarantee therapy success when operating in hard-to-reach body districts, a precise and robust imaging strategy is required for monitoring and controlling MRs in real-time. Ultrasound (US) may represent a powerful technology, but MRs' visibility with US needs to be improved, especially when targeting echogenic tissues. In this context, motions of MRs have been exploited to enhance their contrast, e.g., by Doppler imaging. To exploit a more selective contrast-enhancement mechanism, in this study, we analyze in detail the characteristic motions of one of the most widely adopted MR concepts, i.e., the helical propeller, with a particular focus on its interactions with the backscattered US waves. We combine a kinematic analysis of the propeller 3D motion with an US acoustic phase analysis (APA) performed on the raw radio frequency US data in order to improve imaging and tracking in bio-mimicking environments. We validated our US-APA approach in diverse scenarios, aimed at simulating realistic in vivo conditions, and compared the results to those obtained with standard US Doppler. Overall, our technique provided a precise and stable feedback to visualize and track helical propellers in echogenic tissues (chicken breast), tissue-mimicking phantoms with bifurcated lumina, and in the presence of different motion disturbances (e.g., physiological flows and tissue motions), where standard Doppler showed poor performance. Furthermore, the proposed US-APA technique allowed for real-time estimation of MR velocity, where standard Doppler failed.

Osteoporosis Diagnostic Model Using a Multichannel Convolutional Neural Network Based on Quantitative Ultrasound Radiofrequency Signal.

Quantitative ultrasound (QUS) is a promising screening method for osteoporosis. In this study, a new method to improve the diagnostic accuracy of QUS was established in which a multichannel convolutional neural network (MCNN) processes the raw radiofrequency (RF) signal of QUS. The improvement in the diagnostic accuracy of osteoporosis using this new method was evaluated by comparison with the conventional speed of sound (SOS) method. Dual-energyX-ray absorptiometry was used as the diagnostic standard. After being trained, validated and tested in a data set consisting of 274 participants, the MCNN model could significantly raise the accuracy of osteoporosis diagnosis compared with the SOS method. The adjusted MCNN model performed even better when adjusted by age, height and weight data. The sensitivity, specificity and accuracy of the adjusted MCNN method for osteoporosis diagnosis were 80.86%, 84.23% and 83.05%, respectively; the corresponding values for SOS were 50.60%, 73.68% and 66.67%. The area under the receiver operating characteristic curve of the adjusted MCNN method was also higher than that of SOS (0.846 vs. 0.679). In conclusion, our study indicates that the MCNN method may be more accurate than the conventional SOS method. The MCNN tool and ultrasound RF signal analysis are promising future developmental directions for QUS in screening for osteoporosis.

Compact automatic modulation recognition using over-the-air signals and FOS features

The recent deployment of automatic modulation recognition (AMR) for cognitive radio (CR) systems has significantly enhanced spectrum sensing capabilities. The utilization of real-time over-the-air digital radio frequency (RF) data for the development of a digital spectrum sensing model based on the automatic modulation classification (AMC) is presented in this study as a step for incorporating opportunistic spectrum sensing onto the NomadicBTS architecture. Some digital modulation techniques were studied for second-generation (2G) through fourth-generation (4G) technology. The raw RF signal dataset was digitized and curated, while non-complex first-order statistical (FOS) features were used with algorithms based on the Scaled conjugate gradient (SCG) and Levenberg-Marquardt (LM) to find the best learning algorithm for the generated AMR model. The results show that the developed AMR model has a very high likelihood of correctly classifying signals, with distinct patterns for each of the features of FOS. The results are compared to reveal a least mean square error (MSE) of 0.0131 with a maximum accuracy of 93.5 percent when the model was trained with seventy (70) neurons in the hidden layer using the LM method. The best model's accuracy will allow for the most precise identification of spectrum holes in the bands under consideration.