Ultrasonic Imaging最新文献

Correctly calculating skin stiffness with ultrasound shear wave elastography techniques requires an accurate measurement of skin thickness. We developed and compared two algorithms, a thresholding method and a deep learning method, to measure skin thickness on ultrasound images. Here, we also present a framework for weakly annotating an unlabeled dataset in a time-effective manner to train the deep neural network. Segmentation labels for training were proposed using the thresholding method and validated with visual inspection by a human expert reader. We reduced decision ambiguity by only inspecting segmentations at the center A-line. This weak annotation approach facilitated validation of over 1000 segmentation labels in 2 hours. A lightweight deep neural network that segments entire 2D images was designed and trained on this weakly-labeled dataset. Averaged over six folds of cross-validation, segmentation accuracy was 57% for the thresholding method and 78% for the neural network. In particular, the network was better at finding the distal skin margin, which is the primary challenge for skin segmentation. Both algorithms have been made publicly available to aid future applications in skin characterization and elastography.

Ultrasonic guided wave techniques have been applied to characterize cortical bone for osteoporosis assessment. Compared with the current gold-standard X-ray-based diagnostic methods, ultrasound-based techniques pose some advantages such as compactness, low cost, lack of ionizing radiation, and their ability to detect the mechanical properties of the cortex. Axial transmission technique with a source-receiver offset is employed to acquire the ultrasound data. The dispersion characteristics of the guided waves in bones are normally analyzed in the transformed domains using the dispersion curves. The transformed domain can be time-frequency map using a single channel or wavenumber-frequency (or phase velocity-frequency) map with multi-channels. In terms of acquisition effort, the first method is more cost- and time-effective than the latter. However, it remains unclear whether single-channel dispersion analysis can provide as much quantitative guided-wave information as the multi-channel analysis. The objective of this study is to compare the two methods using numerically simulated and ex vivo data of a simple bovine bone plate and explore their advantages and disadvantages. Both single- and multi-channel signal processing approaches are implemented using sparsity-constrained optimization algorithms to reinforce the focusing power. While the single-channel data acquisition and processing are much faster than those of the multi-channel, modal identification and analysis of the multi-channel data are straightforward and more convincing.

Significant successes in machine learning approaches to image analysis for various applications have energized strong interest in automated diagnostic support systems for medical images. The evolving in-depth understanding of the way carcinogenesis changes the texture of cellular networks of a mass/tumor has been informing such diagnostics systems with use of more suitable image texture features and their extraction methods. Several texture features have been recently applied in discriminating malignant and benign ovarian masses by analysing B-mode images from ultrasound scan of the ovary with different levels of performance. However, comparative performance evaluation of these reported features using common sets of clinically approved images is lacking. This paper presents an empirical evaluation of seven commonly used texture features (histograms, moments of histogram, local binary patterns [256-bin and 59-bin], histograms of oriented gradients, fractal dimensions, and Gabor filter), using a collection of 242 ultrasound scan images of ovarian masses of various pathological characteristics. The evaluation examines not only the effectiveness of classification schemes based on the individual texture features but also the effectiveness of various combinations of these schemes using the simple majority-rule decision level fusion. Trained support vector machine classifiers on the individual texture features without any specific pre-processing, achieve levels of accuracy between 75% and 85% where the seven moments and the 256-bin LBP are at the lower end while the Gabor filter is at the upper end. Combining the classification results of the top k (k = 3, 5, 7) best performing features further improve the overall accuracy to a level between 86% and 90%. These evaluation results demonstrate that each of the investigated image-based texture features provides informative support in distinguishing benign or malignant ovarian masses.

The quality assurance (QA) of ultrasound transducers is often identified as an area requiring continuous development in terms of the tools available to users. Periodic evaluation of the transducers as part of the QA protocol is important, since the quality of the diagnostics. Some of the key criteria determining the process of developing a QA protocol include the complexity of setup, the time required, accuracy, and potential automation to achieve scale. For the current study, a total of eight different ultrasound machines (12 transducers) with linear transducers were obtained separately. The results from these 12 transducers were used to validate the protocol. WAD-QC was used as part of this study to assess in-air reverberation patterns obtained from ultrasound transducers. Initially, three in-air reverberation images obtained from normal transducers and three obtained from defective transducers were used to calculate the uniformity parameters. The results were applied to 12 other images obtained from independent sources. Image processing results with WAD-QC were verified with imageJ. A comparison of raw data for uniformity showed consistency, and using controls based on mean absolute deviation yielded identical results. WAD-QC can be considered as a powerful mechanism for quick, efficient, and accurate analysis of in-air reverberation patterns obtained from ultrasound transducers.

Chronic kidney disease (CKD) is characterized by progressive reductions in skeletal muscle function and size. The concept of muscle quality is increasingly being used to assess muscle health, although the best means of assessment remains unidentified. The use of muscle echogenicity is limited by an inability to be compared across devices. Gray level of co-occurrence matrix (GLCM), a form of image texture analysis, may provide a measure of muscle quality, robust to scanner settings. This study aimed to identify GLCM values from skeletal muscle images in CKD and investigate their association with physical performance and strength (a surrogate of muscle function). Transverse images of the rectus femoris muscle were obtained using B-mode 2D ultrasound imaging. Texture analysis (GLCM) was performed using ImageJ. Five different GLCM features were quantified: energy or angular second moment (ASM), entropy, homogeneity, or inverse difference moment (IDM), correlation, and contrast. Physical function and strength were assessed using tests of handgrip strength, sit to stand-60, gait speed, incremental shuttle walk test, and timed up-and-go. Correlation coefficients between GLCM indices were compared to each objective functional measure. A total of 90 CKD patients (age 64.6 (10.9) years, 44% male, eGFR 33.8 (15.7) mL/minutes/1.73 m²) were included. Better muscle function was largely associated with those values suggestive of greater image texture homogeneity (i.e., greater ASM, correlation, and IDM, lower entropy and contrast). Entropy showed the greatest association across all the functional assessments (r = -.177). All GLCM parameters, a form of higher-order texture analysis, were associated with muscle function, although the largest association as seen with image entropy. Image homogeneity likely indicates lower muscle infiltration of fat and fibrosis. Texture analysis may provide a novel indicator of muscle quality that is robust to changes in scanner settings. Further research is needed to substantiate our findings.

Inaccurate placement of the ablation catheter and the inability to monitor the real-time temperature within the tissue of interest such as veins curbs the treatment efficacy of laser ablation procedures during thermal therapies. Our previous studies have validated the efficacy of photoacoustic (PA) imaging during endovenous laser ablation (EVLA) procedures. However, the PA-guided therapies suffer from low temporal resolution, due to the low pulse repetition rates of pulsed lasers, which could cause a problem during fast catheter motion and rapid temperature changes. Herein, to enhance the accuracy and sensitivity for tracking the ablation catheter tip and temperature monitoring, we proposed to develop a high frame rate (500 Hz), combined ultrasound (US), and PA-guided ablation system. The proposed PA-guided ablation system was evaluated in a set of ex vivo tissue studies. The developed system provides a 2 ms temporal resolution for tracking and monitoring the ablation catheter tip's location and temperature, which is 50 times higher temporal resolution compared to the previously proposed 10 Hz system. The proposed system also provided more accurate feedback about the temperature variations during rapid temperature increments of 10°C per 250 ms. The co-registered US and PA images have an imaging resolution of about 200 μm and a field of view of 45 × 40 mm². Tracking the ablation catheter tip in an excised tissue layer shows higher accuracy during a relatively fast catheter motion (0.5-3 mm/s). The fast US/PA-guided ablation system will potentially enhance the outcome of ablation procedures by providing location and temperature feedback.

Quantitative ultrasound techniques have been previously used to evaluate biological hard tissues, characterized by a large acoustic impedance contrast. Here, we are interested in the imaging of experimental data from different test-targets with high acoustic impedance contrast, using the Full Waveform Inversion (FWI) method to solve the inverse problem. This method is based on high-resolution numerical modeling of the forward problem of interaction between waves and medium, considering the full time series. To reduce the complexity of the numerical implementation, the model considers a fluid medium. Therefore, the aim is to evaluate the precision of the reconstruction under this assumption for materials with a different level of attenuation of shear waves, to study the limits of this hypothesis. Images of the sound speed obtained using the experimental data are presented, and the precision of the reconstruction is evaluated. Future work should include viscoelastic materials.

Spatial frequency analysis (SFA) is a quantitative ultrasound method that characterizes tissue organization. SFA has been used for research involving tendon injury, but may prove useful in similar research involving skeletal muscle. As a first step, we investigated if SFA could detect known architectural differences within hamstring muscles. Ultrasound B-mode images were collected bilaterally at locations corresponding to proximal, mid-belly, and distal thirds along the hamstrings from 10 healthy participants. Images were analyzed in the spatial frequency domain by applying a two-dimensional Fourier Transform in all 6.5 × 6.5 mm kernels in a region of interest corresponding to the central portion of the muscle. SFA parameters (peak spatial frequency radius [PSFR], maximum frequency amplitude [Mmax], sum of frequencies [Sum], and ratio of Mmax to Sum [Mmax%]) were extracted from each muscle location and analyzed by separate linear mixed effects models. Significant differences were observed proximo-distally in PSFR (p = .039), Mmax (p < .0001), and Sum (p < .0001), consistent with architectural descriptions of the hamstring muscles. These results suggest that SFA can detect regional differences of healthy tissue structure within the hamstrings-an important finding for future research in regional muscle structure and mechanics.

In the clinical analysis of Intravascular ultrasound (IVUS) images, the lumen size is an important indicator of coronary atherosclerosis, and is also the premise of coronary artery disease diagnosis and interventional treatment. In this study, a fully automatic method based on deep learning model and handcrafted features is presented for the detection of the lumen borders in IVUS images. First, 193 handcrafted features are extracted from the IVUS images. Then hybrid feature vectors are constructed by combining handcrafted features with 64 high-level features extracted from U-Net. In order to obtain the feature subsets with larger contribution, we employ the extended binary cuckoo search for feature selection. Finally, the selected 36-dimensional hybrid feature subset is used to classify the test images using dictionary learning based on kernel sparse coding. The proposed algorithm is tested on the publicly available dataset and evaluated using three indicators. Through ablation experiments, mean value of the experimental results (Jaccard: 0.88, Hausdorff distance: 0.36, Percentage of the area difference: 0.06) prove to be effective improving lumen border detection. Furthermore, compared with the recent methods used on the same dataset, the proposed method shows good performance and high accuracy.