Search

Breast density is widely adopted to reflect the likelihood of early breast cancer development. Existing methods of mammographic density classification either require steps of manual operations or achieve only moderate classification accuracy due to the limited model capacity. In this study, we present a radiomics approach based on dilated and attention-guided residual learning for the task of mammographic density classification. The proposed method was instantiated with two datasets, one clinical dataset and one publicly available dataset, and classification accuracies of 88.7 and 70.0 percent were obtained, respectively. Although the classification accuracy of the public dataset was lower than the clinical dataset, which was very likely related to the dataset size, our proposed model still achieved a better performance than the naive residual networks and several recently published deep learning-based approaches. Furthermore, we designed a multi-stream network architecture specifically targeting at analyzing the multi-view mammograms. Utilizing the clinical dataset, we validated that multi-view inputs were beneficial to the breast density classification task with an increase of at least 2.0 percent in accuracy and the different views lead to different model classification capacities. Our method has a great potential to be further developed and applied in computer-aided diagnosis systems. Our code is available at https://github.com/lich0031/Mammographic_Density_Classification .

A major advantage of muscle ultrasound is its capacity to capture muscle contraction with decent spatial and temporal resolutions. Ever since the first report of sonomyography, a number of related reports have been put forward by different research groups. Future perspectives of sonomyography, including the development of a multi-modal muscle research platform that integrates ultrasound, EMG, acceleration sensor signals, etc., are discussed in this chapter.

In recent years, with the rapid development of deep learning methods, neural networks have been proven to have great potentials in the field of medical image analysis. Certainly, such technical advancements in artificial intelligence also enable some new possibilities in the investigation of skeletal muscle using ultrasound imaging, both for existing or new topics. In this chapter, two examples related to this advancement are introduced.

The change of muscle thickness (MT) can effectively reflect muscle activity during contraction, and is an important index for musculoskeletal studies in sports science or rehabilitation engineering. This chapter introduces the development of MT measurement technology from multiple perspectives, including: 1D to 2D and cross-sectional to longitudinal plane of muscles. The transverse cross-sectional perspective introduces the application of cross-correlation technology, compression tracking, and other technologies. The longitudinal perspective introduces the methods based on the revoting Hough transform, method based on MI-FFD registration, and local and global intensity fitting (LGIF) method, etc. These technologies developed in the field of computer vision have been successfully applied in musculoskeletal ultrasound and strongly promote the automated process of muscle thickness measurement.

It is well-known that there can be quite serious noises in ultrasound images and ultrasound imaging is sensitive to the orientation of the probe during imaging. Therefore, there are often problems that the image quality is not satisfactory or the image content in the context frames fluctuates greatly, which leads to difficulties in the dynamic assessment of muscles during their contraction. This chapter introduces some research works to enhance ultrasound muscle images, especially one using a deep learning technique.

Recently, an emerging trend in medical image classification is to combine radiomics framework with deep learning classification network in an integrated system. Although this combination is efficient in some tasks, the deep learning-based classification network is often difficult to capture an effective representation of lesion regions, and prone to face the challenge of overfitting, leading to unreliable features and inaccurate results, especially when the sizes of the lesions are small or the training dataset is small. In addition, these combinations mostly lack an effective feature selection mechanism, which makes it difficult to obtain the optimal feature selection. In this paper, we introduce a novel and effective deep semantic segmentation feature-based radiomics (DSFR) framework to overcome the above-mentioned challenges, which consists of two modules: the deep semantic feature extraction module and the feature selection module. Specifically, the extraction module is utilized to extract hierarchical semantic features of the lesions from a trained segmentation network. The feature selection module aims to select the most representative features by using a novel feature similarity adaptation algorithm. Experiments are extensively conducted to evaluate our method in two clinical tasks: the pathological grading prediction in pancreatic neuroendocrine neoplasms (pNENs), and the prediction of thrombolytic therapy efficacy in deep venous thrombosis (DVT). Experimental results on both tasks demonstrate that the proposed method consistently outperforms the state-of-the-art approaches by a large margin.

Accurate and temporal-consistent segmentation of echocardiography is important for diagnosing cardiovascular disease. Existing methods often ignore consistency among the segmentation sequences, leading to poor ejection fraction (EF) estimation. In this paper, we propose to enhance temporal consistency of the segmentation sequences with two co-learning strategies of segmentation and tracking from ultrasonic cardiac sequences where only end diastole and end systole frames are labeled. First, we design an appearance-level co-learning (CLA) strategy to make the segmentation and tracking benefit each other and provide an eligible estimation of cardiac shapes and motion fields. Second, we design another shape-level co-learning (CLS) strategy to further improve segmentation with pseudo labels propagated from the labeled frames and to enforce the temporal consistency by shape tracking across the whole sequence. Experimental results on the largest publicly-available echocardiographic dataset (CAMUS) show the proposed method, denoted as CLAS, outperforms existing methods for segmentation and EF estimation. In particular, CLAS can give segmentations of the whole sequences with high temporal consistency, thus achieves excellent estimation of EF, with Pearson correlation coefficient 0.926 and bias of 0.1%, which is even better than the intra-observer agreement.

Mobile edge caching (MEC) and device to device (D2D) communications are two potential technologies to resolve traffic overload in heterogeneous networks. Prior works usually investigate them separately with MEC for traffic offloading and D2D for information transmission. In this paper, a composite framework consists of MEC and cache-enabled D2D communications is proposed to minimize the energy cost of systematic traffic transmission, where file popularity and user preference are the critical criteria for small base stations (SBSs) and users respectively. Under this framework, we propose a novel caching strategy where Markov decision process (MDP) is applied to model the requesting behaviors of users. A new algorithm based on reinforcement learning (RL) is proposed to reveal the popularity of files as well as users' preference. In particular, Q-learning (QL) algorithm and deep Q- network (DQN) algorithm are respectively applied to users and SBS due to different complexity of status. To save the energy cost of systematic traffic transmission, users acquire partial traffic through D2D communications based on the cached contents. Taking the memory limits, D2D available files and status changing into consideration, the proposed RL algorithm enables users' devices and SBS to prefetch the optimal files while learning, and hence reducing the energy cost significantly. Simulation results demonstrate the superior energy saving performance of the proposed RL-based algorithm over other existing methods under various conditions.

Optical coherence tomography angiography (OCTA) is extensively used for visualizing retinal vasculature, including the foveal avascular zone (FAZ). Assessment of the FAZ is critical in the diagnosis and management of various retinal diseases. Accurately segmenting the FAZ in the deep retinal layer (dFAZ) is very challenging due to unclear capillary terminals. In this study, a customized encoder-decoder deep learning network was used for dFAZ segmentation. Six-fold cross-validation was performed on a total of 80 subjects (63 healthy subjects and 17 diabetic retinopathy subjects). The proposed method obtained an average Dice of 0.88 and an average Hausdorff distance of 17.79, suggesting the dFAZ was accurately segmented. The proposed method is expected to realize good clinical application value by providing an objective and faster and spatially-quantitative preparation of dFAZ-related investigations.

The motion intent of individual finger is recognized based on mechanomyogram.The mechanomyogram signal is captured by inertial sensor.Time and transform domain feature sets are effectively applied for recognition.The effectiveness of single and combined feature sets are evaluated.The best average recognition rate of 95.20% can be achieved. The mechanomyogram (MMG) signals detected from forearm muscle group contain abundant information which can be utilized to predict finger motion intention. Few works have been reported in this area especially for the recognition of individual finger motions, which however is crucial for many applications such as prosthesis control. In this paper, a MMG based finger gesture recognition system is designed to identify the motions of each finger. In this system, three kinds of feature sets, wavelet packet transform (WPT) coefficients, stationary wavelet transform (SWT) coefficients, and the time and frequency domain hybrid (TFDH) features, are adopted and processed by a support vector machine (SVM) classifier. The experimental results show that the average accuracy rates of recognition using the WPT, SWT and TFDH features are 91.64%, 94.31%, and 91.56%, respectively. Furthermore, the average rate of 95.20% can be achieved when above three feature sets are combined to use in the proposed recognition system.

Assessing the location and extent of lesions caused by chronic stroke is critical for medical diagnosis, surgical planning, and prognosis. In recent years, with the rapid development of 2D and 3D convolutional neural networks (CNN), the encoder-decoder structure has shown great potential in the field of medical image segmentation. However, the 2D CNN ignores the 3D information of medical images, while the 3D CNN suffers from high computational resource demands. This paper proposes a new architecture called dimension-fusion-UNet (D-UNet), which combines 2D and 3D convolution innovatively in the encoding stage. The proposed architecture achieves a better segmentation performance than 2D networks, while requiring significantly less computation time in comparison to 3D networks. Furthermore, to alleviate the data imbalance issue between positive and negative samples for the network training, we propose a new loss function called Enhance Mixing Loss (EML). This function adds a weighted focal coefficient and combines two traditional loss functions. The proposed method has been tested on the ATLAS dataset and compared to three state-of-the-art methods. The results demonstrate that the proposed method achieves the best quality performance in terms of DSC = 0.5349 $\pm$ ± 0.2763 and precision = 0.6331 $\pm$ ± 0.295).

An accurate segmentation and quantification of the superficial foveal avascular zone (sFAZ) is important to facilitate the diagnosis and treatment of many retinal diseases, such as diabetic retinopathy and retinal vein occlusion. We proposed a method based on deep learning for the automatic segmentation and quantification of the sFAZ in optical coherence tomography angiography (OCTA) images with robustness to brightness and contrast (B/C) variations. A dataset of 405 OCTA images from 45 participants was acquired with Zeiss Cirrus HD-OCT 5000 and the ground truth (GT) was manually segmented subsequently. A deep learning network with an encoder–decoder architecture was created to classify each pixel into an sFAZ or non-sFAZ class. Subsequently, we applied largest-connected-region extraction and hole-filling to fine-tune the automatic segmentation results. A maximum mean dice similarity coefficient (DSC) of 0.976 ± 0.011 was obtained when the automatic segmentation results were compared against the GT. The correlation coefficient between the area calculated from the automatic segmentation results and that calculated from the GT was 0.997. In all nine parameter groups with various brightness/contrast, all the DSCs of the proposed method were higher than 0.96. The proposed method achieved better performance in the sFAZ segmentation and quantification compared to two previously reported methods. In conclusion, we proposed and successfully verified an automatic sFAZ segmentation and quantification method based on deep learning with robustness to B/C variations. For clinical applications, this is an important progress in creating an automated segmentation and quantification applicable to clinical analysis.

As a medical imaging technology which can show the metabolism of the brain, 18F-fluorodeoxyglucose (FDG)-positron emission tomography (PET) is of great value for the diagnosis of Parkinson's Disease (PD). With the development of pattern recognition technology, analysis of brain images using deep learning are becoming more and more popular. However, existing computer-aided-diagnosis technologies often over fit and have poor generalizability. Therefore, we aimed to improve a framework based on Group Lasso Sparse Deep Belief Network (GLS-DBN) for discriminating PD and normal control (NC) subjects based on FDG-PET imaging. In this study, 225 NC and 125 PD cohorts from Huashan and Wuxi 904 hospitals were selected. They were divided into the training & validation dataset and 2 test datasets. First, in the training & validation set, subjects were randomly partitioned 80:20, with multiple training iterations for the deep learning model. Next, Locally Linear Embedding was used as a dimension reduction algorithm. Then, GLS-DBN was used for feature learning and classification. Different sparse DBN models were used to compare datasets to evaluate the effectiveness of our framework. Accuracy, sensitivity, and specificity were examined to validate the results. Output variables of the network were also correlated with longitudinal changes of rating scales about movement disorders (UPDRS, H&Y). As a result, accuracy of prediction (90% in Test 1, 86% in Test 2) for classification of PD and NC patients outperformed conventional approaches. Output scores of the network were strongly correlated with UPDRS and H&Y (R = 0.705, p < 0.001; R = 0.697, p < 0.001 in Test 1; R = 0.592, p = 0.0018, R = 0.528, p = 0.0067 in Test 2). These results show the GLS-DBN is feasible method for early diagnosis of PD.

High-frame-rate imaging technique, such as Corneal Visualization Scheimpug Technology, plays an important role in evaluating corneal properties. Many studies measure various biomechanical parameters and associate them to different corneal pathological conditions. However, the vertical resolution of the acquired images is limited mainly due to optical distortions and acquisition errors, which in some scenarios (where sufficient vertical resolution is critical) mislead the interpretation and clinical diagnosis of corneal pathological conditions therefore it become critical to establish an efficient way to improve the vertical resolution in Scheimpflug photography. In this work, we propose an effective computational method to retrieve the image details in high-frame-rate imaging, providing both high temporal and vertical resolution. Experiments and a utility assessment including image data from totally 120 subjects on the measurement of central corneal thickness demonstrate a previously-not-recognized pattern of corneal deformation in air puff tests, as well as the statistically significant difference revealed (p

Whether in sports or exercise rehabilitation, muscles are usually strengthened through exercise. In this process, the biological structure of the muscle is also changed subtly. However, it has not reported whether these changes can be reflected in ultrasonic images. Furthermore, it is a challenging and novel task to capture this nuance, and existing methods that use only one image are difficult to resolve it. Therefore, this study proposes a multiple images feature selection (MIFS) framework that combines information from multiple images, aiming to find an optimal feature set that can effectively distinguish between different exercise levels. In this study, the optimal feature set was obtained from seven differently operated musculoskeletal ultrasound images (MUI) of 107 healthy subjects (55 males, 52 females, age 21.0 ± 1.9 years) included 54 regular exercisers and 53 irregular exercisers by the MIFS framework. All ultrasound images are acquired in B-mode with an 8.5 MHz linear array ultrasound transducer. Finally, we obtain an optimal feature set consisting of 20 features, and the optimal classification accuracy exceeds the existing single-image-based method, up to 78.9 %. This preliminary study suggests nuance in muscle structure caused by different exercise levels can be captured by ultrasound images. What's more, MIFS can more accurately deal with complicated task of classifying between regular and irregular exercisers as well.

This paper proposes a segmentation-free radiomics method to classify malignant and benign breast tumors with shear-wave elastography (SWE) data. The method is targeted to integrate the advantage of both SWE in providing important elastic with morphology information and convolutional neural network (CNN) in automatic feature extraction and accurate classification. Compared to traditional methods, the proposed method is designed to directly extract features from the dataset without the prerequisite of segmentation and manual operation. This can keep the peri-tumor information, which is lost by segmentation-based methods. With the proposed model trained on 540 images (318 of malignant breast tumors and 222 of benign breast tumors, respectively), an accuracy of 95.8%, a sensitivity of 96.2%, and a specificity of 95.7% was obtained for the final test. The superior performances compared to the existing state-of-the-art methods and its automatic nature both demonstrate that the proposed method has a great potential to be applied to clinical computer-aided diagnosis of breast cancer.

I258 the study of frequency modulated continuous wave (FMCW) radar, it is difficult to complete an experiment with application-specific radars, since adapting their parameters bring much cost and time-consuming work. To address this problem, this paper presents an efficient and automated FMCW radar experimental platform. As a control center, a computer is integrated with a programmable transmitter, a programmable receiver and RF components in the platform. This platform can flexibly configure the parameters of the FMCW signal and realize automatic transmission, reception and signal processing in an automated and efficient way. The effectiveness of the platform is demonstrated by a laboratory experiment.

Motion-intent-based finger gesture recognition systems are crucial for many applications such as prosthesis control, sign language recognition, wearable rehabilitation system, and human–computer interaction. In this article, a motion-intent-based finger gesture recognition system is designed to correctly identify the tapping of every finger for the first time. Two auto-event annotation algorithms are firstly applied and evaluated for detecting the finger tapping frame. Based on the truncated signals, the Wavelet packet transform (WPT) coefficients are calculated and compressed as the features, followed by a feature selection method that is able to improve the performance by optimizing the feature set. Finally, three popular classifiers including naive Bayes (NBC), K-nearest neighbor (KNN), and support vector machine (SVM) are applied and evaluated. The recognition accuracy can be achieved up to 94%. The design and the architecture of the system are presented with full system characterization results.

Unilateral peripheral facial paralysis (UPFP) is a form of facial nerve paralysis and clinically classified according to conditions of facial symmetry. Prompt and precise assessment is crucial to neural rehabilitation of UPFP. The prevalent House–Brackmann (HB) grading system relies on subjective judgments with significant interobservation variation. Therefore, to explore an objective method for the UPFP assessment, clinical image sequences are captured using a web camera setup while 5 healthy and 27 UPFP subjects perform a group of predefined actions, including keeping expressionless, raising brows, closing eyes, bulging cheek, and showing teeth in turn. First, facial region is decided using Haar cascade classifier, and then landmark points are acquired by a supervised descent method. Second, these landmark points are used to generate a group of features reflecting the structural parameters of regions of eyebrows, eyes, nose, and mouth, respectively. Third, correlation coefficients are computed between the raw features HB scores. To reduce feature dimensions, only those with correlation coefficients larger than an empirically selected value, 0.35, are input into a support vector machine to generate a classifier. With the classifier, exact match (discrepancy = 0 between result from proposed method and HB scores) rate at 49.9%, and loose match (discrepancy = 1) rate at 87.97% are achieved on the experiment data. After sample augmentation, the final rate is increased to 90.01%, outperformed previous reports. In conclusion, it is demonstrated with an unobtrusive web camera setup, encouraging results have been generated with the proposed framework in this exploratory study.

Unilateral peripheral facial paralysis (UPFP) is a form of facial nerve paralysis and clinically classified according to facial asymmetry. Prompt and precise assessment is crucial to the neural rehabilitation of UPFP. For UPFP assessment, most of the existing assessment systems are subjective and empirical. Therefore, an objective assessment system will help clinical doctors to obtain a prompt and precise assessment. Distinguishing precisely between degrees of asymmetry is hard using pure pattern recognition methods. Thus, a novel objective assessment process based on convolutional neuronal networks is proposed in this paper that provides an end-to-end solution. This method could alleviate the problem and produced a classification accuracy of 91.25% for predicting the House-Brackmann degree on a given UPFP image dataset.

Muscle thickness measurement in ultrasonography was traditionally conducted by a trained operator, and the manual detecting process is time consuming and subjective. In this paper, we proposed an automatic tracking strategy to achieve the continuous and quantitative measurement for gastrocnemius muscle thickness in ultrasound images. The method involved three steps: tracking of seed points, contours extraction of aponeuroses, and muscle thickness estimation. In an ultrasound image sequence, we first selected two seed points in the first frame manually for the superficial and deep aponeuroses, respectively. Seed points in all following frames were then tracked by registering to their respective previous frames. Second, we adopted the local and global intensity fitting model to extract the contours of aponeuroses. At last, the muscle thickness was achieved by calculating the distance between the contours of superficial and deep aponeuroses. The performance of the algorithm was evaluated using 500 frames of ultrasound images. It was demonstrated in the experiments that the proposed methods could be used for objective tracking of aponeuroses and estimation of muscle thickness in musculoskeletal ultrasound images.

Ultrasonography is an important diagnostic imaging technique for visualization of tendons, which provides useful health diagnostic and fundamental information in neuromuscular studies of human motion systems. Conventional ultrasonic-based tendon studies, however, are highly dependent on subjective experience of operators due to various impairments of ultrasound images. Dynamic changes of muscle and tendon deformation in a sequence can hardly be manually processed. Consequently, there is an urgent need for automatic analysis of tendon behavior. This paper proposes an automatic ultrasonic tendon tracking algorithm to extract the shape deformation of central tendon of rectus femoris (CT-RF) from ultrasonic image sequences. The tracking problem is complicated by the highly deformable tendon, time-varying brightness, and the inconspicuousness of the target. To address this difficult tracking problem, we proposed a new intensity-compensated free-form deformation (IC-FFD)-based tracking algorithm with local shape refinement (LSR). Experimental results and comparison show that the proposed IC-FFD-LSR algorithm outperforms IC-FFD and conventional methods such as MI-FFD in CT-RF tracking.

Ultrasound imaging has been used frequently for the study of muscle contraction, including measurements of pennation angles and fascicle orientations. However, these measurements were traditionally conducted by manually drawing lines on the ultrasound images. In this study, we proposed a modified Hough transform (HT), aiming at automatically estimating orientations of straight line-shaped patterns, such as muscle fibers and muscle-bone interface in ultrasound images. The new method first located the global maximum in the HT accumulator matrix, which corresponded to the most dominant collinear feature points globally, using the standard HT; then the pixels close to the detected line were removed from the edge map, the HT accumulator matrix was calculated again, i.e., revoting, and a new line was detected. The iteration was repeated until the predefined termination conditions were satisfied. The performance of the algorithm was tested using computer-generated images with different levels of noises, as well as clinical ultrasound images, and compared with that of the conventional method. It was found that the orientation estimation results obtained by the new algorithm were well correlated (R2 = 0.965), with those obtained using the traditional method, i.e., drawing lines manually and reading the angles with the assistance of software. Further mean-difference plots revealed a difference of 0.18 +/- 2.41 degrees between the two methods at the 95% confidence level. Compared with the traditional method, the new algorithm was more capable of handling with highly noisy data and could avoid the aliasing problem, i.e., reporting multiple lines instead of single expected line. The results of this study suggested that the proposed revoting HT can be potentially used for the reliable and nonsubjective automatic estimation of the orientations of muscle fibers in musculoskeletal ultrasound images.

A new method is reported for monitoring respiratory activity using photoplethysmography (PPG) without the need of prior knowledge of the respiratory rate range. Two channels of transmission mode PPG signals were collected from the subject's index finger and analyzed using an independent component analysis (ICA) algorithm. The respiratory activity was separated from the heart‐related pulsation in PPG after the ICA analysis. The results demonstrated that the ICA analysis could successfully extract normal and simulated apnoea respiratory activities. The algorithm was applied for both simulated signals and those collected from 10 young normal adults. The reported technique may be potentially used for the simultaneous monitoring of the pulse rate and respiratory activity based on the hardware of current pulse oximetry devices. The structure and implementation of this preliminary respiratory activity monitoring system were presented. Possible further improvements for the system performance were also di...

An efficient image registration algorithm, the Inverse Compositional image alignment method based on minimization of Sum of Squared Differences of images, is applied in fundus blood vessel angiography under confocal scanning laser ophthalmoscope, to build image mosaics which have larger field of view without loss of resolution to assist diagnosis. Furthermore, based on similar technique, the angiography video stabilization algorithm is implemented for fundus documenting. The actual underlying models of motion between images and corresponding convergence criteria are also discussed. The experiment results in fundus images demonstrate the effectiveness of the registration scheme.

Currently, the evaluation of placental maturity has mainly focused on subjective measure, which highly depends on the observation and experiences of the clinicians and not reliable. This paper proposes a new method for grading placenta maturity in B-mod ultrasound (US) images automatically based on local intensity order pattern (LIOP) and fisher vector (FV). After extracting invariant LIOP feature from the affine covariant region, the feature is encoded by FV to improve the classification accuracy and reduce the processing time. Experimental results show the effectiveness of the proposed method with an accuracy of 0.9375, a sensitivity of 0.9804 and a specificity of 0.9375 for the placental maturity grading. Moreover, experimental results demonstrate that the LIOP feature outperforms the traditional SIFT feature for grading.

As a direct determinant parameter to quantify muscle activity, the muscle thickness (MT) has been investigated in many aspects and for various purposes. Ultrasonography (US) is a promising modality to detect muscle morphological changes during contractions since it is portable, noninvasive, and real time. However, there are few reports on sensitive and efficient estimation of changes of MT in a cross-sectional plane. In this feasibility investigation, we proposed a coarse-to-fine method based on a compressive-tracking algorithm for estimation of MT changes during an example task of isometric knee extension using ultrasound images. The sensitivity and efficiency are evaluated with 1920 US images from quadriceps muscle (QM) in eight subjects. The detection results were compared with those obtained from both traditional manual measurement and the well known normalized cross-correlation method, and the effect of the size of tracking window on detection performance was evaluated as well. It is demonstrated that the proposed method agrees well with the manual measurement. Meanwhile, it is not only sensitive to relatively small changes of MT but also computationally efficient.

As an important parameter related to musculoskeletal functions, muscle thickness has been studied for various purposes. However, muscle thickness is usually measured manually by an experienced clinical expert, which is subjective and time consuming, and there are few studies on automatic tracking of muscle thickness during dynamic contraction. In this paper, we proposed a modified Hough transform (HT) to achieve the quantitative and continuous measurement for muscle thickness in ultrasound images. The method involved three steps: image enhancement, locating of superficial and deep aponeuroses by RVHT, and computation of the distance between aponeuroses. The performance of the new method is evaluated using ultrasound images from gastrocnemius muscles of seven patients. The result from the proposed method is also compared to manual detection and another method which was based on Compressive Tracking Algorithm (CTA) applied in our previous work. It was demonstrated in the experiment that the proposed method agrees well with the manual measurement and was able to provide a more convenient and effective approach than the CTA. It could be used for objective muscle thickness tracking in musculoskeletal ultrasound images.

Muscle thickness is an important parameter related to musculoskeletal functions and has been studied in many ways and for various purposes. In recent years, ultrasound imaging has been widely used for measuring muscle properties of human muscles non-invasively, with the advantages of real time and low cost. The muscle thickness is usually measured manually, which is subjective and time consuming. In addition, there are few studies on automatic estimation of muscle thickness during dynamic contraction. In this study, an automatic estimation method based on compressive tracking algorithm is proposed to detect the thickness changes of tibialis anterior muscle during dynamic contraction on ultrasound images. The performance of the proposed method is compared to manual detection using clinical images from tibialis anterior muscles of ten patients. As a result, we found that the proposed method agrees well with the manual measurement, and it was able to provide an accurate and efficient approach for estimating muscle thickness during human motion.

We built a Placido’s disc videokeratographic system intended for eye examination of small animals. To derive the corneal topology and other corneal biometric parameters we rely on precise extraction of Placido ring patterns. However, since the images from CCD are noisy, image enhancement is necessary as a pre-processing procedure in order to prevent further noise magnification during the corneal surface topography reconstruction. In this study, two popular image enhancement methods, namely Gabor Filtering (GF) and Multiscale Vessel Enhancement Filtering (MVEF), are quantitatively evaluated in the context of image enhancement before corneal reconstruction. The two methods were tested on 21 images of steel balls with different diameters. Both methods have been demonstrated to be useful for Placido ring extraction; however, the GF method showed better performance on images of steel balls with smaller diameter. Compared to the GF method, the MVEF method performed better on images of steel balls with larger diameter. The result of the present work implies that the two methods should be used cooperatively.

In recent years, ultrasound imaging have been widely used to assess human muscle. Quantitative ultrasound can provide objective information about muscle architectures and changes of motion pattern during muscle contraction. However, there are few reports about the visualization of cross-sectional skeletal muscle motion using descriptive color coding strategy. In this study, ultrasound images were recorded from right quadriceps femoris muscle during isometric knee extension, and a motion estimation method based on a theory for warping was used to compute morphological changes of muscle during contraction. Then a novel color coding strategy, which agrees with the established directing rules in medical imaging, was applied to present the motion field. In comparison with the traditional color coding method, the new strategy was more capable of providing detailed information about the velocity and morphological changes in different region of muscle intuitively. The proposed approach has the potential to be applied for the functional understanding and segmentation of various muscle sections.

Background Muscle fiber orientation (MFO) is an important parameter related to musculoskeletal functions. The traditional manual method for MFO estimation in sonograms was labor-intensive. The automatic methods proposed in recent years also involved voting procedures which were computationally expensive. Methods In this paper, we proposed a new framework to efficiently estimate MFO in sonograms. We firstly employed Multi-scale Vessel Enhancement Filtering (MVEF) to enhance fascicles in the sonograms and then the enhanced images were binarized. Finally, line-shaped patterns in the binary map were detected one by one, according to their shape properties. Specifically speaking, for the long-and-thinner regions, the orientation of the targeted muscle fibre was directly computed, without voting procedures, as the orientation of the ellipse that had the same normalized second central moments as the region. For other cases, the Hough voting procedure might be employed for orientation estimation. The performance of the algorithm was evaluated using four various group of sonograms, which are a dataset used in previous reports, 33 sonograms of gastrocnemius from 11 young healthy subjects, one sonogram sequence including 200 frames from a subject and 256 frames from an aged subject with cerebral infarction respectively. Results It was demonstrated in the experiments that measurements of the proposed method agreed well with those of the manual method and achieved much more efficiency than the previous Re-voting Hough Transform (RVHT) algorithm. Conclusions Results of the experiments suggested that, without compromising the accuracy, in the proposed framework the previous orientation estimation algorithm was accelerated by reduction of its dependence on voting procedures.

Skeletal muscle is an important tissue of human body, and its contractions control and regulate body motions. Muscle contraction results in morphological changes of the related muscles. Ultrasound imaging is an effective tool for studying muscle architectures and monitoring the morphological changes of muscles. The latter process can be realized with a motion estimation algorithm. However, ultrasound images are usually corrupted by speckle noises and performance of motion estimation methods can be significantly affected by the noises. To get a better performance in motion analysis, in this paper, as a pre-processing step, an adaptive filter named adaptive guided image filtering (AGF) is suggested to reduce speckle noises. We first transformed the multiplicative noise model into an additive one by taking the logarithm of the original speckled data, then performed AGF to obtain the filtered image, and finally took the tackled image back into exponent. Experimental results showed that AGF had a better performance in terms of noise attenuation and edge preservation compared with other standard filters. In quantitative results, the filtered images also had the highest Peak-Signal-to-Noise Ratio (PSNR) using AGF. It's believed that AGF is a good choice for the pre-processing stage of muscle motion analysis.

Ultrasonography is a convenient and widely used technique to look into the longitudinal muscle motion as it is radiation-free and real-time. The motion of localized parts of the muscle, disclosed by ultrasonography, spatially reflects contraction activities of the corresponding muscles. However, little attention was paid to the estimation of longitudinal muscle motion, especially towards estimation of dense deformation field at different depths under the skin. Yet fewer studies on the visualization of such muscle motion or further clinical applications were reported in the literature. A primal–dual algorithm was used to estimate the motion of gastrocnemius muscle (GM) in longitudinal direction in this study. To provide insights into the rules of longitudinal muscle motion, we proposed a novel framework including motion estimation, visualization and quantitative analysis to interpret synchronous activities of collaborating muscles with spatial details. The proposed methods were evaluated on ultrasound image sequences, captured at a rate of 25 frames per second from eight healthy subjects. In order to estimate and visualize the GM motion in longitudinal direction, each subject was asked to perform isometric plantar flexion twice. Preliminary results show that the proposed visualization methods provide both spatial and temporal details and they are helpful to study muscle contractions. One of the proposed quantitative measures was also tested on a patient with unilateral limb dysfunction caused by cerebral infarction. The measure revealed distinct patterns between the normal and the dysfunctional lower limb. The proposed framework and its associated quantitative measures could potentially be used to complement electromyography (EMG) and torque signals in functional assessment of skeletal muscles.

Ultrasonography is a convenient and widely-used technique to look into the muscle contraction as it is non-invasive and real-time. Muscle thickness has been investigated in many aspects and for various purposes. However, there are few studies on automatic estimation of changes of muscle thickness in cross-sectional plane during contraction. In this study we proposed a coarse-to-fine method based on compressive tracking algorithm for real-time estimation of muscle thickness during isometric contraction on ultrasound images. The real-time performance and precision are investigated with data from quadriceps muscle in five subjects. The detection results were compared with those obtained from both cross-correlation algorithm and manual measurement. It is shown that the proposed method agrees well with the manual measurement and outperforms the cross-correlation method in the sense of both accuracy and computation cost.

Ultrasonography is a convenient and widely-used technique to view the longitudinal muscle motion as it is non-invasive and real-time. However, the estimation of longitudinal muscle motion, in sense of dense deformation field, was paid little attention to previously. In this report a primal-dual algorithm was used to estimate the motion of gastrocnemius muscle (GM) in longitudinal direction. Then a color encoding scheme and the mean velocities (mm/s) of frame-wise motion field were adopted to show the synchronous activities. The preliminary results, on 600 frames of ultrasound muscle images in motion, demonstrated that the proposed estimation method are helpful to study of the contraction of muscle in longitudinal direction, providing motion details both spatially and temporally.

In this paper, we applied the Adaptive Non-local Filtering to reduce the speckle noises and strengthen the meaningful features in musculoskeletal sonograms, then the structural features points were extracted after binarization, and finally we applied Re-voting Hough transform (RVHT) method proposed earlier to estimate the muscle fiber orientations. The results obtained by this new method was compared with those obtained using a previous method using the Sobel edge operator and RVHT only. 226 lines representing the orientation of muscle fibers or boundaries were manually selected by two operators from 45 muscular ultrasound images. These lines were also detected by the proposed automatic estimation methods. Using the results obtained manually as a reference, the differences of the new method and that using RVHT only were 0.40± 0.44 degrees and 0.70 ± 0.73 degrees, respectively. The results demonstrated that the performance of the orientation estimation method with the additional image processing approaches was fairly improved with few extra computation costs.

Ultrasound elastography has become a wellknown optional imaging method for the diagnosis of tissue abnormalities in various body parts. It images the elasticity of compliant tissues by estimating the local displacements and strains using pre- and post-compression RF echo signals. In this paper, taking the RF signal as image intensity and RF samples as pixels, we present a motion estimation framework to compute the axial tissue displacements and strains. This method takes advantage of both the block matching algorithm (BMA) and local optical flow techniques. For two frames of RF signals, coarse motion estimates are first computed using BMA. The motion estimates obtained are then used to warp the first frame toward the second one, thus making the warped frame more spatially correlated to the second one. Next, the Lucas-Kanade optical flow method is employed to compute the residual motion between the warped frame and the original second frame, with inherent sub-pixel precision. Finally, the displacements from the two steps are combined. The warp-and-refine procedure can be iterated if the residual motion is larger than a predefined empirical threshold. To test its feasibility, we first applied the method to simulated data. The results show that our method is robust to relatively large motions and is capable of generating accurate motion estimation with subsample spatial resolution. These methods have been deployed and are being tested on a commercialized ultrasound machine that previously did not have elastography functions. Quality real-time display of elastography along with freehand scanning has been accomplished. The proposed framework provides an alternative method for motion estimation with good performance, and it can potentially be improved using hardware to realize the BMA.

Ultrasound imaging has been widely used to investigate the morphological changes during skeletal muscle contraction. In this study, the ultrasound images were recorded from extensor muscle during finger flexion, and the optical flow algorithm was used to recognize the muscle deformation for different fingers' flexions. The preliminary results demonstrated that the directions of optical flow and deformation areas of muscle contraction at certain location were different for five fingers' flexions. The proposed approach has the potential to classify finger flexion with the feature extraction from optical flow field.

In this study, the Gabor filter bank technique was applied to the biceps and gastrocnemius ultrasound images to longitudinally enhance the coherently oriented and hyperechoic perimysiums regions. The method involved three steps, orientation field estimation, frequency map computation and Gabor filtering. The method was evaluated using a simulated image distorted with multiplicative speckle noises, where the “muscles” were arranged in a bipennate fashion with a central “aponeurosis”. After the enhancement using the proposed method, most of the original hyperechoic bands in the simulated image could be recovered and the noises in other locations were greatly reduced. The proposed method was also tested on biceps sonograms collected from 4 healthy adult subjects. Based on the filtering results, the hyperechoic regions, especially the fibroadipose septas, could be enhanced in several aspects, such as the completion of broken but coherently oriented regions. It is believed that the proposed method have potentials in assisting the visualization of strongly oriented patterns in ultrasound images and especially the quantitative estimation of muscle thickness, muscle fiber pennation angle and fascicle length.

In this paper, the traditional snake model and gradient vector flow (GVF) snake model are studied, which are believed to be quite slow due to the need to compute inverse matrix. Actually, the GVF in the latter snake model is formed by a biased linear diffusion procedure, and there would be oscillations around the edge of the object. Based on GVF generated through non-linear diffusion, we present a fast GVF (FGVF) snake model which is much faster than the traditional snake model and GVF snake model, and would cause no degradation of stability and flexibility, meanwhile, it could reduce the oscillations around the edges. The segmentation results using FGVF and error analysis on simulated images are presented. Finally, the demonstration of FGVF applied to Computed Tomography and Magnetic Resonance images are shown, the segmentation results are satisfactory visually with much less computation time in comparison with former snakes.

Background Ultrasonography is a convenient technique to investigate muscle properties and has been widely used to look into muscle functions since it is non-invasive and real-time. Muscle thickness, a quantification which can effectively reflect the muscle activities during muscle contraction, is an important measure for musculoskeletal studies using ultrasonography. The traditional manual operation to read muscle thickness is subjective and time-consuming, therefore a number of studies have focused on the automatic estimation of muscle fascicle orientation and muscle thickness, to which the speckle noises in ultrasound images could be the major obstacle. There have been two popular methods proposed to enhance the hyperechoic regions over the speckles in ultrasonography, namely Gabor Filtering and Multiscale Vessel Enhancement Filtering (MVEF). Methods A study on gastrocnemius muscle is conducted to quantitatively evaluate whether and how these two methods could help the automatic estimation of the muscle thickness based on Revoting Hough Transform (RVHT). The muscle thickness results obtained from each of the two methods are compared with the results from manual measurement, respectively. Data from an aged subject with cerebral infarction is also studied. Results It’s shown in the experiments that, Gabor Filtering and MVEF can both enable RVHT to generate comparable results of muscle thickness to those by manual drawing (mean ± SD, 1.45 ± 0.48 and 1.38 ± 0.56 mm respectively). However, the MVEF method requires much less computation than Gabor Filtering. Conclusions Both methods, as preprocessing procedure can enable RVHT the automatic estimation of muscle thickness and MVEF is believed to be a better choice for real-time applications.