Congzhi Wang

Deep learning based classification of breast tumors with shear-wave elastography.

This study aims to build a deep learning (DL) architecture for automated extraction of learned-from-data image features from the shear-wave elastography (SWE), and to evaluate the DL architecture in differentiation between benign and malignant breast tumors. We construct a two-layer DL architecture for SWE feature extraction, comprised of the point-wise gated Boltzmann machine (PGBM) and the restricted Boltzmann machine (RBM). The PGBM contains task-relevant and task-irrelevant hidden units, and the task-relevant units are connected to the RBM. Experimental evaluation was performed with five-fold cross validation on a set of 227 SWE images, 135 of benign tumors and 92 of malignant tumors, from 121 patients. The features learned with our DL architecture were compared with the statistical features quantifying image intensity and texture. Results showed that the DL features achieved better classification performance with an accuracy of 93.4%, a sensitivity of 88.6%, a specificity of 97.1%, and an area under the receiver operating characteristic curve of 0.947. The DL-based method integrates feature learning with feature selection on SWE. It may be potentially used in clinical computer-aided diagnosis of breast cancer.

Quantification of Elastic Heterogeneity Using Contourlet-Based Texture Analysis in Shear-Wave Elastography for Breast Tumor Classification

Ultrasound shear-wave elastography (SWE) has become a valuable tool for diagnosis of breast tumors. The purpose of this study was to quantify the elastic heterogeneity of breast tumors in SWE by using contourlet-based texture features and evaluating their diagnostic performance for classification of benign and malignant breast tumors, with pathologic results as the gold standard. A total of 161 breast tumors in 125 women who underwent B-mode and SWE ultrasonography before biopsy were included. Five quantitative texture features in SWE images were extracted from the directional subbands after the contourlet transform, including the mean (Tmean), maximum (Tmax), median (Tmed), third quartile (Tqt), and standard deviation (Tsd) of the subbands. Diagnostic performance of the texture features and the classic features was compared using the area under the receiver operating characteristic curve (AUC) and the leave-one-out cross validation with Fisher classifier. The feature Tmean achieved the highest AUC (0.968) among all features and it yielded a sensitivity of 89.1%, a specificity of 94.3% and an accuracy of 92.5% for differentiation between benign and malignant tumors via the leave-one-out cross validation. Compared with the best classic feature, i.e., the maximum elasticity, Tmean improved the AUC, sensitivity, specificity and accuracy by 3.5%, 12.7%, 2.8% and 6.2%, respectively. The Tmed, Tqt and Tsd were also superior to the classic features in terms of the AUC and accuracy. The results demonstrated that the contourlet-based texture features captured the tumors elastic heterogeneity and improved diagnostic performance contrasted with the classic features.

Improved Anatomical Specificity of Non-invasive Neuro-stimulation by High Frequency (5 MHz) Ultrasound.

Low frequency ultrasound (<1 MHz) has been demonstrated to be a promising approach for non-invasive neuro-stimulation. However, the focal width is limited to be half centimeter scale. Minimizing the stimulation region with higher frequency ultrasound will provide a great opportunity to expand its application. This study first time examines the feasibility of using high frequency (5 MHz) ultrasound to achieve neuro-stimulation in brain, and verifies the anatomical specificity of neuro-stimulation in vivo. 1 MHz and 5 MHz ultrasound stimulation were evaluated in the same group of mice. Electromyography (EMG) collected from tail muscles together with the motion response videos were analyzed for evaluating the stimulation effects. Our results indicate that 5 MHz ultrasound can successfully achieve neuro-stimulation. The equivalent diameter (ED) of the stimulation region with 5 MHz ultrasound (0.29 ± 0.08 mm) is significantly smaller than that with 1 MHz (0.83 ± 0.11 mm). The response latency of 5 MHz ultrasound (45 ± 31 ms) is also shorter than that of 1 MHz ultrasound (208 ± 111 ms). Consequently, high frequency (5 MHz) ultrasound can successfully activate the brain circuits in mice. It provides a smaller stimulation region, which offers improved anatomical specificity for neuro-stimulation in a non-invasive manner.

Computer-assisted assessment of ultrasound real-time elastography: Initial experience in 145 breast lesions

PURPOSE To develop and evaluate a computer-assisted method of quantifying five-point elasticity scoring system based on ultrasound real-time elastography (RTE), for classifying benign and malignant breast lesions, with pathologic results as the reference standard. MATERIALS AND METHODS Conventional ultrasonography (US) and RTE images of 145 breast lesions (67 malignant, 78 benign) were performed in this study. Each lesion was automatically contoured on the B-mode image by the level set method and mapped on the RTE image. The relative elasticity value of each pixel was reconstructed and classified into hard or soft by the fuzzy c-means clustering method. According to the hardness degree inside lesion and its surrounding tissue, the elasticity score of the RTE image was computed in an automatic way. Visual assessments of the radiologists were used for comparing the diagnostic performance. Histopathologic examination was used as the reference standard. The Students t test and receiver operating characteristic (ROC) curve analysis were performed for statistical analysis. RESULTS Considering score 4 or higher as test positive for malignancy, the diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were 93.8% (136/145), 92.5% (62/67), 94.9% (74/78), 93.9% (62/66), and 93.7% (74/79) for the computer-assisted scheme, and 89.7% (130/145), 85.1% (57/67), 93.6% (73/78), 92.0% (57/62), and 88.0% (73/83) for manual assessment. Area under ROC curve (Az value) for the proposed method was higher than the Az value for visual assessment (0.96 vs. 0.93). CONCLUSION Computer-assisted quantification of classical five-point scoring system can significantly eliminate the interobserver variability and thereby improve the diagnostic confidence of classifying the breast lesions to avoid unnecessary biopsy.

Extraction of Respiratory Activity from Photoplethysmographic Signals Based on an Independent Component Analysis Technique: Preliminary Report

Abstract 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 subjects 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 discussed.

Shear Modulus Estimation on Vastus Intermedius of Elderly and Young Females over the Entire Range of Isometric Contraction

Elderly people often suffer from sarcopenia in their lower extremities, which gives rise to the increased susceptibility of fall. Comparing the mechanical properties of the knee extensor/flexors on elderly and young subjects is helpful in understanding the underlying mechanisms of the muscle aging process. However, although the stiffness of skeletal muscle has been proved to be positively correlated to its non-fatiguing contraction intensity by some existing methods, this conclusion has not been verified above 50% maximum voluntary contraction (MVC) due to the limitation of their measurement range. In this study, a vibro-ultrasound system was set up to achieve a considerably larger measurement range on muscle stiffness estimation. Its feasibility was verified on self-made silicone phantoms by comparing with the mechanical indentation method. The system was then used to assess the stiffness of vastus intermedius (VI), one of the knee extensors, on 10 healthy elderly female subjects (56.7±4.9 yr) and 10 healthy young female subjects (27.6±5.0 yr). The VI stiffness in its action direction was confirmed to be positively correlated to the % MVC level (R2 = 0.999) over the entire range of isometric contraction, i.e. from 0% MVC (relaxed state) to 100% MVC. Furthermore, it was shown that there was no significant difference between the mean VI shear modulus of the elderly and young subjects in a relaxed state (p>0.1). However, when performing step isometric contraction, the VI stiffness of young female subjects was found to be larger than that of elderly participants (p<0.001), especially at the relatively higher contraction levels. The results expanded our knowledge on the mechanical property of the elderly’s skeletal muscle and its relationship with intensity of active contraction. Furthermore, the vibro-ultrasound system has a potential to become a powerful tool for investigating the elderly’s muscle diseases.

Comparison between reflection-mode photoplethysmography and arterial diameter change detected by ultrasound at the region of radial artery.

ObjectiveContinuous monitoring of arterial blood pressure (ABP) can provide vital information on the cardiovascular system in clinical practice and research. To achieve this, methods have been developed to determine ABP continuously and noninvasively, including the wrist volume clamp method, which uses reflection-mode photoplethysmography (PPG) to indicate the volume change of the radial artery. However, this kind of indication is reportedly not accurate enough, and an in-vivo test for its accuracy is still lacking. Our aim was to fix this gap. MethodsAs ultrasound can detect the arterial diameter accurately; we have developed a sensor comprised a PPG sensor and an ultrasound transducer, which can collect reflection-mode PPG and A-mode ultrasound signals, simultaneously. Tests on 15 volunteers were conducted under the similar conditions of using wrist volume clamp method for a comparison between the PPG signal and the radial arterial diameter. ResultsIt was observed that there were obvious differences between the shapes of these two signals, mainly between their drifting trend parts, with the pulsatile components of the signals matching very well. The mean root mean square value of the differences of the trend parts (0.287±0.072) was found to be much larger than that of the pulsatile parts (0.107±0.028). ConclusionThe results indicated that using PPG signal as a reference for ABP measurement at the wrist region might not be accurate enough, and ultrasound has the potential for replacing it in the wrist volume clamp method for more accurate ABP determination.

Strain estimation in elastography using scale-invariant keypoints tracking

This paper proposes a novel strain estimator using scale-invariant keypoints tracking (SIKT) for ultrasonic elastography. This method is based on tracking stable features between the pre- and post-compression A-lines to obtain tissue displacement estimates. The proposed features, termed scaleinvariant keypoints, are independent of signal scale change according to the scale-space theory, and therefore can preserve their patterns while undergoing a substantial range of compression. The keypoints can be produced by searching for repeatedly assigned points across all possible scales constructed from the convolution with a one-parameter family of Gaussian kernels. Because of the distinctive property of the keypoints, the SIKT method could provide a reliable tracking over changing strains, an effective resistance to anamorphic noise and sonographic noise, and a significant reduction in processing time. Simulation and experimental results show that the SIKT method is able to provide better sensitivity, a larger dynamic range of the strain filter, higher resolution, and a better contrast- to-noise ratio (CNRe) than the conventional methods. Moreover, the computation time of the SIKT method is approximately 5 times that of the cross-correlation techniques.

Home-Telecare of the elderly living alone using an new designed ear-wearable sensor

This paper presents a new ear-wearable vital sign sensor designed for homecare of the elderly or other vulnerable people living alone. The device was comprised of a sensing body equipped with a pair of infrared LED and phototransistor and a dual-axis accelerometer, and an MP3 device-like control box with batteries. The sensing body was worn on the subjectpsilas ear and could detect the photoplethysmography (PPG) signals from the back of the subjectpsilas auricle. The PPG signals were sent to a computer wirelessly for heart rate calculation and abnormality detection. The program could automatically alarm and transmit the saved data to a center through modem when the heart rate was out of the normal range which decided by a time-wise baseline algorithm. The acceleration signals were used to eliminate the motion artifacts in the calculation of the heart rate. The comparison of the results obtained by this ear-wearable sensor and a commercially available fingertip sensor showed that the accuracy and reliability of it were acceptable. It was also found in the field test that the ear-wearable sensor could be worn comfortably and inconspicuously by the subject and the system could work continuously for approximately 120 hours without replacement of the batteries.

Development of a Mechanical Scanning Device With High-Frequency Ultrasound Transducer for Ultrasonic Capsule Endoscopy

Wireless capsule endoscopy has opened a new era by enabling remote diagnostic assessment of the gastrointestinal tract in a painless procedure. Video capsule endoscopy is currently commercially available worldwide. However, it is limited to visualization of superficial tissue. Ultrasound (US) imaging is a complementary solution as it is capable of acquiring transmural information from the tissue wall. This paper presents a mechanical scanning device incorporating a high-frequency transducer specifically as a proof of concept for US capsule endoscopy (USCE), providing information that may usefully assist future research. A rotary solenoid-coil-based motor was employed to rotate the US transducer with sectional electronic control. A set of gears was used to convert the sectional rotation to circular rotation. A single-element focused US transducer with 39-MHz center frequency was used for high-resolution US imaging, connected to an imaging platform for pulse generation and image processing. Key parameters of US imaging for USCE applications were evaluated. Wire phantom imaging and tissue phantom imaging have been conducted to evaluate the performance of the proposed method. A porcine small intestine specimen was also used for imaging evaluation in vitro. Test results demonstrate that the proposed device and rotation mechanism are able to offer good image resolution (