Weiqi Wang

Doppler ultrasound signal denoising based on wavelet frames

A novel approach was proposed to denoise the Doppler ultrasound signal. Using this method, wavelet coefficients of the Doppler signal at multiple scales were first obtained using the discrete wavelet frame analysis. Then, a soft thresholding-based denoising algorithm was employed to deal with these coefficients to get the denoised signal. In the simulation experiments, the SNR improvements and the maximum frequency estimation precision were studied for the denoised signal. From the simulation and clinical studies, it was concluded that the performance of this discrete wavelet frame (DWF) approach is higher than that of the standard (critically sampled) wavelet transform (DWT) for the Doppler ultrasound signal denoising.

Measurement of the Dispersion and Attenuation of Cylindrical Ultrasonic Guided Waves in Long Bone

Osteoporotic bones are likely to have less cortical bone than healthy bones. The velocities of guided waves propagating in a long cylindrical bone are very sensitive to bone properties and cortical thickness (CTh). This work studies the dispersion and attenuation of ultrasonic guided waves propagating in long cylindrical bone. A hollow cylinder filled with a viscous liquid was used to model the long bone and then to calculate the theoretical phase and group velocities, as well as the attenuation of the waves. The generation and selection of guided wave modes were based on theoretical dispersive curves. The phase velocity and attenuation of cylindrical guided wave modes, such as L(0,1), L(0,2) and L(0,3), were measured in bovine tibia using angled beam transducers at various propagation distances ranging from 75 to 160 mm. The results showed that the phase velocity of the L(0,2) guided wave mode decreased with an increase in CTh. The attenuation of the low cylindrical guided wave modes was a nonlinear function that increased with propagation distance and mode order. The L(0,2) mode had a different attenuation for each CTh. The experimental results were in good agreement with the predicted values. Cylindrical guided waves of low-frequency and low-order have been shown to demonstrate more dispersion and less attenuation and should, therefore, be used to evaluate long bone.

Mode separation of Lamb waves based on dispersion compensation method.

Ultrasonic Lamb modes typically propagate as a combination of multiple dispersive wave packets. Frequency components of each mode distribute widely in time domain due to dispersion and it is very challenging to separate individual modes by traditional signal processing methods. In the present study, a method of dispersion compensation is proposed for the purpose of mode separation. This numerical method compensates, i.e., compresses, the individual dispersive waveforms into temporal pulses, which thereby become nearly un-overlapped in time and frequency and can thus be extracted individually by rectangular time windows. It was further illustrated that the dispersion compensation also provided a method for predicting the plate thickness. Finally, based on reversibility of the numerical compensation method, an artificial dispersion technique was used to restore the original waveform of each mode from the separated compensated pulse. Performances of the compensation separation techniques were evaluated by processing synthetic and experimental signals which consisted of multiple Lamb modes with high dispersion. Individual modes were extracted with good accordance with the original waveforms and theoretical predictions.

Multiridge-based analysis for separating individual modes from multimodal guided wave signals in long bones

Quantitative ultrasound has great potential for assessing human bone quality. Considered as an elastic waveguide, long bone supports propagation of several guided modes, most of which carry useful information, individually, on different aspects of long bone properties. Therefore, precise knowledge of the behavior of each mode, such as velocity, attenuation, and amplitude, is important for bone quality assessment. However, because of the complicated characteristics of the guided waves, including dispersion and mode conversion, the measured signal often contains multiple wave modes, which yields the problem of mode separation. In this paper, some novel signal processing approaches were introduced to solve this problem. First, a crazy-climber algorithm was used to separate time-frequency ridges of individual modes from time-frequency representations (TFR) of multimodal signals. Next, corresponding time domain signals representing individual modes were reconstructed from the TFR ridges. It was found that the separated TFR ridges were in agreement with the theoretical dispersion, and the reconstructed signals were highly representative of the individual guided modes as well. The validations of this study were analyzed by simulated multimodal signals, with or without noise, and by in vitro experiments. Results of this study suggest that the ridge detection and individual reconstruction method are suitable for separating individual modes from multimodal signals. Such a method can improve the analysis of skeletal guided wave signals by providing accurate assessment of mode-specific ultrasonic parameters, such as group velocity, and indicate different bone quality properties.

Spectrogram enhancement algorithm: a soft thresholding-based approach

Enhancing the spectrogram by denoising the Doppler ultrasound signal is a preliminary step, and important for further processing. Because the spectrogram may be based on the short-time fast Fourier transform (FFT) of the Doppler ultrasound signal, whose power spectrum density is time-varying, traditional denoising algorithms that simply optimize the mean-squared error are not appropriate, and they may exhibit considerable undesirable, noise-induced frequency components. A soft thresholding-based denoising algorithm is put forward in this paper, that achieves almost the minimax mean square error (MSE) over a wide range of function classes having norms measuring smoothness (i.e., it meets both the requirement of smoothness and MSE). Due to the importance of noise level estimation while applying this method, several robust L-estimators are compared and the median absolute deviation (MAD) method is chosen to estimate the noise level. The simulation study shows better performance of the later algorithm under various quantification measures, compared to the FFT thresholding and the hard thresholding wavelet method, and the results of clinical data also confirm it.

Analysis of frequency dependence of ultrasonic backscatter coefficient in cancellous bone

The ultrasonic scattering mechanism in cancellous bone is investigated theoretically and a model describing the frequency dependence of ultrasonic scattering from cancellous bone is presented. The ultrasonic backscatter coefficient (BSC) of bovine tibiae, human calcanei in vitro and in vivo, were measured and discussed. The data of BSC were also fitted by polynomial. The results demonstrate that BSC is a nonlinear function of frequency and increases with frequency. A good agreement was obtained between BSC values from theory and experiment. Also, the high correlation coefficient between BSC and bone mineral density was obtained, r=0.85+/-0.07 (mean+/-SD) (n=15, p<0.001). Based on the values of BSC, the status of cancellous bone and the degree of osteoporotic fracture risk may be assessed.

Estimating coronary artery lumen area with optimization-based contour detection

A modified optimization-based contour detection method was presented to compute the lumen area of the coronary artery from intravascular ultrasound (IVUS) video images. First, the search range for the artery inner wall was determined based on the continuity of IVUS video frames. Next, the internal and external energy were calculated to describe the smoothness of the arterial wall and the grayscale variation of ultrasound images, respectively. Here, a novel form of the external energy which combines the gradient and variance of the intensity of image in the radial direction was used. Finally, the minimal energy path based on the optimum contour of the artery wall was obtained using circular dynamic programming (DP). By the comparison with the typical DP procedure using the traditional external energy form, based only on the image gradient, the reliability of this modified method is considerably improved in the measurement of coronary artery lumen area.

Analysis of Superimposed Ultrasonic Guided Waves in Long Bones by the Joint Approximate Diagonalization of Eigen-matrices Algorithm

The parameters of ultrasonic guided waves (GWs) are very sensitive to mechanical and structural changes in long cortical bones. However, it is a challenge to obtain the group velocity and other parameters of GWs because of the presence of mixed multiple modes. This paper proposes a blind identification algorithm using the joint approximate diagonalization of eigen-matrices (JADE) and applies it to the separation of superimposed GWs in long bones. For the simulation case, the velocity of the single mode was calculated after separation. A strong agreement was obtained between the estimated velocity and the theoretical expectation. For the experiments in bovine long bones, by using the calculated velocity and a theoretical model, the cortical thickness (CTh) was obtained. For comparison with the JADE approach, an adaptive Gaussian chirplet time-frequency (ACGTF) method was also used to estimate the CTh. The results showed that the mean error of the CTh acquired by the JADE approach was 4.3%, which was smaller than that of the ACGTF method (13.6%). This suggested that the JADE algorithm may be used to separate the superimposed GWs and that the JADE algorithm could potentially be used to evaluate long bones.

Analysis of Apparent Integrated Backscatter Coefficient and Backscattered Spectral Centroid Shift in Calcaneus in vivo for the Ultrasonic Evaluation of Osteoporosis

The purposes of our study were to evaluate the correlation among apparent integrated backscatter coefficient (AIB), spectral centroid shift (SCS) of ultrasonic backscatter signals and bone mineral density (BMD) and to examine the effectiveness of ultrasound variables as predictors of osteoporosis. A total of 1011 persons aged 21-80 y old were included. All study participants underwent BMD measurements of the lumbar spine (LSBMD) and the femoral neck (FNBMD). The participants also underwent calcaneal measurements to determine AIB and SCS with central frequencies of 3.5 (one transducer) and 5.0 MHz (the other transducer). AIB decreased with age and was positively correlated with BMD, while SCS increased with age and was negatively correlated with BMD. The correlation coefficient of SCS with LSBMD and FNBMD at 3.5 MHz was -0.72 and -0.70, respectively. The correlation coefficient at 5.0 MHz was -0.75 and -0.74, respectively. The correlation coefficient of AIB with LSBMD and FNBMD at 3.5 MHz was 0.65 and 0.63. The correlation coefficient at 5.0 MHz was 0.59 and 0.55, respectively. The correlation between SCS and BMD was significantly better than the correlation between AIB and BMD. Using receiver operating characteristic analysis, a significant difference was found between the areas under the curve for SCS and AIB at 3.5 MHz (0.781 vs. 0.715, respectively, p < 0.05), as well as at 5.0 MHz (0.782 vs. 0.709, respectively, p < 0.05). The optimum T-score threshold for SCS was -1.3 for both transducers. The sensitivity and specificity of SCS at 3.5 MHz and 5.0 MHz for the optimum threshold were 64%, 85%, 63% and 86%, respectively. In conclusion, the correlations among the ultrasound parameters and BMDs are strong. SCS performs better than AIB in differentiating patients with osteoporosis. Ultrasound variables may be taken into consideration as predictors of osteoporosis in the future considering its portability.

Automatic segmentation of calcifications in intravascular ultrasound images using snakes and the contourlet transform.

It is valuable to detect calcifications in intravascular ultrasound images for studies of coronary artery diseases. An image segmentation method based on snakes and the Contourlet transform is proposed to automatically and accurately detect calcifications. With the Contourlet transform, an original image is decomposed into low-pass bands and band-pass directional sub-bands. The 2-D Renyis entropy is used to adaptively threshold the low-pass bands in a multiresolution hierarchy to determine regions-of-interest (ROIs). Then a mean intensity ratio, reflecting acoustic shadowing, is presented to classify calcifications from noncalcifications and obtain initial contours of calcifications. The anisotropic diffusion is used in bandpass directional sub-bands to suppress noise and preserve calcific edges. Finally, the contour deformation in the boundary vector field is used to obtain final contours of calcifications. The method was evaluated via 60 simulated images and 86 in vivo images. It outperformed a recently proposed method, the Santos Filho method, by 2.76% and 14.53%, in terms of the sensitivity and specificity of calcification detection, respectively. The area under the receiver operating characteristic curve increased by 0.041. The relative mean distance error, relative difference degree, relative arc difference, relative thickness difference and relative length difference were reduced by 5.73%, 19.79%, 11.62%, 12.06% and 20.51%, respectively. These results reveal that the proposed method can automatically and accurately detect calcifications and delineate their boundaries. (E-mail: yywang@fudan.edu.cn).