A new high resolution color flow system using an eigendecomposition-based adaptive filter for clutter rejection

We present a new signal processing strategy for high frequency color flow mapping in moving tissue environments. A new application of an eigendecomposition-based clutter rejection filter is presented with modifications to deal with high blood-to-clutter ratios (BCR). Additionally, a new method for correcting blood velocity estimates with an estimated tissue motion profile is detailed. The performance of the clutter filter and velocity estimation strategies is quantified using a new swept-scan signal model. In vivo color flow images are presented to illustrate the potential of the system for mapping blood flow in the microcirculation with external tissue motion.     

Clutter filters adapted to tissue motion in ultrasound color flow imaging

The quality of ultrasound color flow images is highly dependent on sufficient attenuation of the clutter signals originating from stationary and slowly moving tissue. Without sufficient clutter rejection, the detection of low velocity blood flow will be poor, and the velocity estimates will have a large bias. In some situations, e.g., when imaging the coronary arteries or when the operator moves the probe in search for small vessels, there is considerable movement of tissue. It has been suggested that clutter rejection can be improved by mixing down the signal with an estimate of the mean frequency prior to high pass filtering. In this paper, we compare this algorithm with several other adaptive clutter filtering algorithms using both experimental data and simulations. We found that realistic accelerations of the tissue have a large effect on the clutter rejection. The best results were obtained by mixing down the signal with non-constant phase increments estimated from the signal. This adapted the filter to a possibly accelerated tissue motion and produced a significant improvement in clutter rejection     

Single-ensemble-based eigen-processing methods for color flow imaging--Part I. The Hankel-SVD filter

Because of their adaptability to the slow-time signal contents, eigen-based filters have shown potential in improving the flow detection performance of color flow images. This paper proposes a new eigen-based filter called the Hankel-SVD filter that is intended to process each slowtime ensemble individually. The new filter is derived using the notion of principal Hankel component analysis, and it achieves clutter suppression by retaining only the principal components whose order is greater than the clutter eigen-space dimension estimated from a frequency based analysis algorithm. To assess its efficacy, the Hankel-SVD filter was first applied to synthetic slow-time data (ensemble size: 10) simulated from two different sets of flow parameters that model: 1) arterial imaging (blood velocity: 0 to 38.5 cm/s, tissue motion: up to 2 mm/s, transmit frequency: 5 MHz, pulse repetition period: 0.4 ms) and 2) deep vessel imaging (blood velocity: 0 to 19.2 cm/s, tissue motion: up to 2 cm/s, transmit frequency: 2 MHz, pulse repetition period: 2.0 ms). In the simulation analysis, the post-filter clutter-to- blood signal ratio (CBR) was computed as a function of blood velocity. Results show that for the same effective stopband size (50 Hz), the Hankel-SVD filter has a narrower transition region in the post-filter CBR curve than that of another type of adaptive filter called the clutter-downmixing filter. The practical efficacy of the proposed filter was tested by application to in vivo color flow data obtained from the human carotid arteries (transmit frequency: 4 MHz, pulse repetition period: 0.333 ms, ensemble size: 10). The resulting power images show that the Hankel-SVD filter can better distinguish between blood and moving-tissue regions (about 9 dB separation in power) than the clutter-downmixing filter and a fixed-rank multi ensemble-based eigen-filter (which showed a 2 to 3 dB separation).     

Optimal velocity estimation in ultrasound color flow imaging in presence of clutter

In color flow imaging (CFI), the rejection of tissue clutter signal is treated separately from blood velocity estimation by high-pass filtering the received Doppler signal. The complete suppression of clutter is then difficult to achieve without affecting the subsequent velocity estimates. In this work, a different approach to velocity estimation is investigated, based on a statistical model of the signal from both clutter and blood. An analytic expression for the Cramer-Rao lower bound (CRLB) is developed, and used to determine the existence of an efficient maximum likelihood estimator (MLE) of blood velocity in CFI when assuming full knowledge of the clutter statistics. We further simulate and compare the performance of the MLE to that of the autocorrelation method (ACM) using finite-impulse response (FIR) and polynomial regression clutter filters. Two signal scenarios are simulated, representing a central and peripheral vessel. Simulations showed that, by including 3-9 (independent) spatial points, the MLE variance approached the CRLB in both scenarios. The ACM was approximately unbiased only for the central scenario in the clutter filter pass band, then with a variance of up to four times the CRLB. The ACM suffered from a severe bias in the filter transition region, and a significant performance gain was achieved here using the MLE. For practical use, the clutter properties must be estimated. We finally replaced the known clutter statistics with an estimate obtained from low-rank approximations of the received sample correlation matrix. Used in the model-based framework, this method came close to the performance of the MLE, and it may be an important step toward a practical model-based estimator, including tissue clutter with optimal performance.     

Clutter filter design for ultrasound color flow imaging

For ultrasound color flow images with high quality, it is important to suppress the clutter signals originating from stationary and slowly moving tissue sufficiently. Without sufficient clutter rejection, low velocity blood flow cannot be measured, and estimates of higher velocities will have a large bias. The small number of samples available (8 to 16) makes clutter filtering in color flow imaging a challenging problem. In this paper, we review and analyze three classes of filters: finite impulse response (FIR), infinite impulse response (IIR), and regression filters. The quality of the filters was assessed based on the frequency response, as well as on the bias and variance of a mean blood velocity estimator using an autocorrelation technique. For FIR filters, the frequency response was improved by allowing a non-linear phase response. By estimating the mean blood flow velocity from two vectors filtered in the forward and backward direction, respectively, the standard deviation was significantly lower with a minimum phase filter than with a linear phase filter. For IIR filters applied to short signals, the transient part of the output signal is important. We analyzed zero, step, and projection initialization, and found that projection initialization gave the best filters. For regression filters, polynomial basis functions provide effective clutter suppression. The best filters from each of the three classes gave comparable bias and variance of the mean blood velocity estimates. However, polynomial regression filters and projection-initialized IIR filters had a slightly better frequency response than could be obtained with FIR filters     

Adaptive clutter rejection filtering in ultrasonic strain-flow imaging

This paper introduces strain-flow imaging as a potential new technique for investigating vascular dynamics and tumor biology. The deformation of tissues surrounding pulsatile vessels and the velocity of fluid in the vessel are estimated from the same data set. The success of the approach depends on the performance of a digital filter that must separate echo signal components caused by flow from tissue motion components that vary spatially and temporally. Eigenfilters, which are an important tool for naturally separating signal components adaptively throughout the image, perform very well for this task. The method is examined using two tissue-mimicking flow phantoms that provide stationary and moving clutter associated with pulsatile flow.     

Real-time adaptive clutter rejection filtering in color flow imaging using power method iterations

We propose a new algorithm for real-time, adaptive-clutter-rejection filtering in ultrasound color flow imaging (CFI) and related techniques. The algorithm is based on regression filtering using eigenvectors of the signal correlation matrix as a basis for representing clutter, a method that previously has been considered too computationally demanding for real-time processing in general CFI applications. The data acquisition and processing scheme introduced allows for a more localized sampling of the clutter statistics and, therefore, an improved clutter attenuation for lower filter orders. By using the iterative power method technique, the dominant eigenvalues and corresponding eigenvectors of the correlation matrix can be estimated efficiently, rendering real-time operation feasible on desktop computers. A new adaptive filter order algorithm is proposed that successfully estimates the proper dimension of the clutter basis, previously one of the major drawbacks of this clutter-rejection technique. The filter algorithm performance and computational demands has been compared to that of conventional clutter filters. Examples have been included which confirms that, by adapting the clutter-rejection filter to estimates of the clutter-signal statistics, improved attenuation of the clutter signal can be achieved in normal as well as more excessive cases of tissue movement and acceleration.     

Single-ensemble-based eigen-processing methods for color flow imaging--Part II. The matrix pencil estimator

Parametric spectral estimators can potentially be used to obtain flow estimates directly from raw slow-time ensembles whose clutter has not been suppressed. We present a new eigen-based parametric flow estimation method called the matrix pencil, whose principles are based on a matrix form under the same name. The presented method models the slow-time signal as a sum of dominant complex sinusoids in the slow-time ensemble, and it computes the principal Doppler frequencies by using a generalized eigen-value problem-formulation and matrix rank reduction principles. Both fixed rank (rank-one, rank-two) and adaptive-rank matrix pencil flow estimators are proposed, and their potential applicability to color flow signal processing is discussed. For the adaptive-rank estimator, the nominal rank was defined as the minimum eigen-structure rank that yields principal frequency estimates with a spread greater than a prescribed bandwidth. In our initial performance evaluation, the fixed-rank matrix pencil estimators were applied to raw color flow data (transmit frequency: 5 MHz; pulse repetition period: 0.175 ms; ensemble size: 14) acquired from a steady flow phantom (70 cm/s at centerline) that was surrounded by rigid-tissue-mimicking material. These fixed-rank estimators produced velocity maps that are well correlated with the theoretical flow profile (correlation coefficient: 0.964 to 0.975). To facilitate further evaluation, the matrix pencil estimators were applied to synthetic slow-time data (transmit frequency: 5 MHz; pulse repetition period: 1.0 ms; ensemble size: 10) modeling flow scenarios without and with tissue motion (up to 1 cm/s). The bias and root-mean-squared error of the estimators were computed as a function of blood-signal-to-noise ratio and blood velocity. The matrix pencil flow estimators showed that they are comparatively less biased than most of the existing frequency-based flow estimators like the lagone autocorrelator.     

Graphics Processing Unit-Based High-Frame-Rate Color Doppler Ultrasound Processing

Color Doppler ultrasound is a routinely used diagnostic tool for assessing blood flow information in real time. The required signal processing is computationally intensive, involving autocorrelation, linear filtering, median filtering, and thresholding. Because of the large amount of data and high computational requirement, color Doppler signal processing has been mainly implemented on custom-designed hardware, with software-based implementation - particularly on a general- purpose CPU - not being successful. In this paper, we describe the use of a graphics processing unit for implementing signal-processing algorithms for color Doppler ultrasound that achieves a frame rate of 160 fps for frames comprising 500 scan lines times 128 range samples, with each scan line being obtained from an ensemble size of 8 with an 8-tap FIR clutter filter.     

A new time-domain narrowband velocity estimation technique for Doppler ultrasound flow imaging. II. Comparative performance assessment

For pt.I see ibid., vol.45, no.4, pp.939-54 (1998). The statistical performance of the new 2-D narrowband time-domain root-MUSIC blood velocity estimator described previously is evaluated using both simulated and flow phantom wideband (50% fractional bandwidth) ultrasonic data. Comparisons are made with the standard 1-D Kasai estimator and two other wideband strategies: the time domain correlator and the wideband point maximum likelihood estimator. A special case of the root-MUSIC, the "spatial" Kasai, is also considered. Simulation and flow phantom results indicate that the root-MUSIC blood velocity estimator displays a superior ability to reconstruct spatial blood velocity information under a wide range of operating conditions. The root-MUSIC mode velocity estimator can be extended to effectively remove the clutter component from the sample volume data. A bimodal velocity estimator is formed by processing the signal subspace spanned by the eigenvectors corresponding to the two largest eigenvalues of the Doppler correlation matrix. To test this scheme, in vivo common carotid flow complex Doppler data was obtained from a commercially available color flow imaging system. Velocity estimates were made using a reduced form of this data corresponding to higher frame rates. The extended root-MUSIC approach was found to produce superior results when compared to both 1- and 2-D Kasai-type estimators that used initialized clutter filters. The results obtained using simulated, flow phantom, and in vivo data suggest that increased sensitivity as well as effective clutter suppression can be achieved using the root-MUSIC technique, and this may be particularly important for wideband high frame rate imaging applications.     

Adaptive clutter rejection for ultrasound color flow imaging based on recursive eigendecomposition

In the conventional eigenfilter used to reject clutter components of ultrasound color flow imaging, input samples are required to be statistically stationary. However, clutter movements may vary over the depth of the imaged area, which makes the eigenfilter less efficient. In the current study, a novel clutter rejection method is proposed based on the recursive eigendecomposition algorithm. In this method, the current eigenvector matrix of the ultrasound echo correlation matrix, which will be used to construct the clutter subspace, is determined by previous eigenvector matrices and the current input. After the estimated clutter signal is obtained by projecting the original input into the clutter space, each filtered output is eventually obtained by subtracting the estimated clutter signal from the original input. This procedure is iterated for each sample volume along the depth. During the updating process, a forgetting factor is introduced to determine proper weights for different inputs. Simulated data in 3 situations and in vivo data collected from human carotid arteries are used to compare the proposed method with other popular clutter filters. Results show that the proposed method can achieve the most accurate velocity profiles in all simulation situations and introduces the fewest velocity artifacts in the tissue region in the in vivo experiment.     

Ultrasound three-dimensional velocity measurements by featuretracking

This article describes a new angle-independent method suitable for three-dimensional (3-D) blood flow velocity measurement that tracks features of the ultrasonic speckle produced by a pulse echo system. In this method, a feature is identified and followed over time to detect motion. Other blood flow velocity measurement methods typically estimate velocity using one- (1-D) or two-dimensional (2-D) spatial and time information. Speckle decorrelation due to motion in the elevation dimension may hinder this estimate of the true 3-D blood flow velocity vector. Feature tracking is a 3-D method with the ability to measure the true blood velocity vector rather than a projection onto a line or plane. Off-line experiments using a tissue phantom and a real-time volumetric ultrasound imaging system have shown that the local maximum detected value of the speckle signal may be identified and tracked for measuring velocities typical of human blood flow. The limitations of feature tracking, including the uncertainty of the peak location and the duration of the local maxima are discussed. An analysis of the expected error using this method is given     

Blood flow imaging--A new real-time, 2-D flow imaging technique

This paper presents a new method for the visualization of two-dimensional (2-D) blood flow in ultrasound imaging systems called blood flow imaging (BFI). Conventional methods of color flow imaging (CFI) and power Doppler (PD) techniques are limited as the velocity component transversal to the ultrasound beam cannot be estimated from the received Doppler signal. The BFI relies on the preservation and display of the speckle pattern originating from the blood flow scatterer signal, and it provides qualitative information of the blood flow distribution and movement in any direction of the image. By displaying speckle pattern images acquired with a high frame rate in slow motion, the blood flow movement can be visually tracked from frame to frame. The BFI is easily combined with conventional CFI and PD methods, and the resulting display modes have been shown to have several advantages compared to CFI or PD methods alone. Two different display modes have been implemented: one combining BFI with conventional CFI, and one combining BFI with PD. Initial clinical trials have been performed to assess the clinical usefulness of BFI. The method especially has potential in vascular imaging, but it also shows potential in other clinical applications.     

Three-stage approach to ultrasound contrast detection

A new method for detecting ultrasound contrast agents using a three-stage pulsing sequence is proposed. The method is based on observations showing that the scattering properties of contrast agents are modified by ultrasonic insonation at high power, but remain unchanged at low power. The objective of the first stage of the pulsing sequence is to use low power pulses to obtain a high resolution reference image without altering the agent. Higher power pulses in the second stage modify the contrast agent. The third stage detects the changes imposed to the contrast agent using low power pulses. A temporal filter is proposed to discriminate contrast response from clutter signal. The method is similar to power Doppler methods in that it uses several pulses to survey the target while destroying the agent. The new idea is to separate detection and destruction to circumvent a trade-off between sensitivity and resolution. Results from in vitro experiments with three different contrast agents are presented. The results are compared with harmonic power Doppler processed from the same data and show that an improvement in sensitivity is achievable by including the high power burst in the pulsing sequence. The results also show that the proposed filter reduces clutter artifacts from moving tissue     

Clutter rejection filters in color flow imaging: a theoretical approach

A general class of linear clutter rejection filters is described, covering the commonly used filter types including FIR/IIR filters with linear initialization, as well as regression filters, where the clutter component is estimated by least square curve fitting. The filter can be described by a complex valued matrix, and a frequency response is defined. However, in contrast to a time invariant filter, the general linear filter may create frequency components which are not present in the input signal. This produces bias in the velocity and velocity spread estimates. It is shown that the clutter filter effect on the autocorrelation estimates can be described by a frequency domain transfer function, but unlike time invariant filters, the transfer function is different for each temporal lag of the autocorrelation function. Using a two dimensional (axial and temporal dimension) model of the received signal, the bias in velocity and velocity spread is quantified, both for the autocorrelation algorithm and the time shift cross-correlation estimator. Theoretical expressions, as well as numerical examples are given.     

On the performance of regression and step-initialized IIR clutterfilters for color Doppler systems in diagnostic medical ultrasound

One of the major issues in color Doppler ultrasound is the suppression of clutter that arises from stationary or slowly moving tissue. If not adequately suppressed, clutter can severely affect the ability of color Doppler systems to accurately estimate the Doppler mean frequency and power of blood, resulting in a potentially misleading depiction of flow. In this study, the performances of two classes of clutter suppression techniques-step-initialized infinite impulse response (IIR) and regression filters-were evaluated and compared by means of extensive simulations. The performance indicators used were the accuracy and precision of the mean frequency and the power estimates after clutter filtering. In summary, the ability of both filter classes to suppress clutter was found to vary considerably depending on factors such as the clutter-to-flow-signal ratio and the ensemble length. In particular, the performance of step-initialized IIR filters was shown to be noticeably inferior to that predicted by their steady-state response. Regression filters were found to offer significantly better performance than step-initialized IIR filters under heavy clutter conditions and, given their steeper roll-off, appear to be more effective clutter suppressors for power Doppler imaging. However, it should be noted that, as demonstrated by the simulations, the performance of IIR filters is severely degraded by their transient response which, in turn, is determined by the initialization scheme used. Therefore, more elaborate schemes-with superior transient characteristics than step-initialization-could significantly improve the effectiveness of IIR filtering under heavy clutter conditions     

In-vivo synthetic aperture flow imaging in medical ultrasound

A new method for acquiring flow images using synthetic aperture techniques in medical ultrasound is presented. The new approach makes it possible to have a continuous acquisition of flow data throughout the whole image simultaneously, and this can significantly improve blood velocity estimation. Any type of filter can be used for discrimination between tissue and blood flow without initialization, and the number of lines used for velocity estimation is limited only by the nonstationarity of the flow. The new approach is investigated through both simulations and measurements. A flow rig is used for generating a parabolic laminar flow, and a research scanner is used for acquiring RF data from individual transducer elements. A reference profile is calculated from a mass flow meter. The parabolic velocity profile is estimated using the new approach with a relative standard deviation of 2.2% and a mean relative bias of 3.4% using 24 pulse emissions at a flow angle of 45 degrees. The 24 emissions can be used for making a full-color flow map image. An in-vivo image of flow in the carotid artery for a 29-year-old male also is presented. The full image is acquired using 24 emissions.     

Ultrasound doppler measurements of low velocity blood flow:limitations due to clutter signals from vibrating muscles

Skeletal muscles vibrate under sustained contraction, and generate low frequency side band clutter in the doppler signal. Both shivering in the hand of the operator and muscle vibrations in the patient itself give rise to the clutter. Clutter rejection filters are commonly used to remove the low frequency components, but the doppler signal from low velocity blood flow is then also lost. This paper describes a model for the pulsed wave (PW) doppler signal from vibrating muscles, reviews a model for the PW doppler signal from moving blood, and by comparing these models presents a theoretical minimum for detectable blood velocity in small vessels, being typically 6.4 mm/s for 6 MHz doppler. The limit has a nonlinear relation to the ultrasound frequency. The model also shows that the radial component of the muscle vibrations can be estimated from the phase of the doppler signal     

Maximum likelihood blood velocity estimator incorporating properties of flow physics

The aspect of correlation among the blood velocities in time and space has not received much attention in previous blood velocity estimators. The theory of fluid mechanics predicts this property of the blood flow. Additionally, most estimators based on a cross-correlation analysis are limited on the maximum velocity detectable. This is due to the occurrence of multiple peaks in the cross-correlation function. In this study a new estimator (CMLE), which is based on correlation (C) properties inherited from fluid flow and maximum likelihood estimation (MLE), is derived and evaluated on a set of simulated and in vivo data from the carotid artery. The estimator is meant for two-dimensional (2-D) color flow imaging. The resulting mathematical relation for the estimator consists of two terms. The first term performs a cross-correlation analysis on the signal segment in the radio frequency (RF)-data under investigation. The flow physic properties are exploited in the second term, as the range of velocity values investigated in the cross-correlation analysis are compared to the velocity estimates in the temporal and spatial neighborhood of the signal segment under investigation. The new estimator has been compared to the cross-correlation (CC) estimator and the previously developed maximum likelihood estimator (MLE). The results show that the CMLE can handle a larger velocity search range and is capable of estimating even low velocity levels from tissue motion. The CC and the MLE produce incorrect velocity estimates due to the multiple peaks, when the velocity search range is increased above the maximum detectable velocity. The root-mean square error (RMS) on the velocity estimates for the simulated data is on the order of 7 cm/s (14%) for the CMLE, and it is comparable to the RMS for the CC and the MLE. When the velocity search range is set to twice the limit of the CC and the MLE, the number of incorrect velocity estimates are 0, 19.1, and 7.2% for the CMLE, CC, and MLE, respectively. The ability to handle a larger search range and estimating low velocity levels was confirmed on in vivo data.     

An extended autocorrelation method for estimation of blood velocity

The conventional autocorrelation method (AM) to estimate the blood velocity for color flow imaging (CFI) is based on the phase estimation of the autocorrelation function. In this paper, a new extended autocorrelation method (EAM) that uses both phase and magnitude of the two dimensional (depth and temporal direction) autocorrelation function for estimating the blood velocity is presented. It is shown that the EAM has similar performance as the cross-correlation method (CCM). Both of them have smaller estimation variance than the AM and have the ability to estimate velocities beyond the Nyquist velocity, but the EAM is more computationally efficient than the CCM. A 2-D blood flow signal with rectilinear velocity including the transit time effect has also been simulated and the results are presented in this paper. For comparison, the EAM and the CCM have been applied to the simulated signals in which the flow velocities are up to four times the Nyquist velocity. The EAM has been further verified by experimental RF data from the subclavian artery