The use of the wavelet transform to describe embolic signals.

A number of methods to detect cerebral emboli and differentiate them from artefacts using Doppler ultrasound have been described in the literature. In most, Fourier transform-based (FT) spectral analysis has been used. The FT is not ideally suited to analysis of short-duration embolic signals due to an inherent trade-off between temporal and frequency resolution. An alternative approach that might be expected to describe embolic signals well is the wavelet transform. Wavelets are ideally suited for the analysis of sudden short-duration signal changes. Therefore, we have implemented a wavelet-based analysis and compared the results of this with a conventional FFT-based analysis. The temporal resolution, as measured by the half-width maximum, was significantly better for the continuous wavelet transform (CWT), mean (SD) 8.40 (8.82) ms, compared with the 128-point FFT, 12.92 (9.70) ms, and 64-point FFT, 10.80 (5.69) ms. Time localization of the CWT for the embolic signal was also significantly better than the FFT. The wavelet transform appears well suited to the analysis of embolic signals offering superior time resolution and time localization to the FFT.     

A real-time autoregressive spectrum analyzer for Doppler ultrasound signals

A system based on a digital signal processor and a microcomputer has been programmed to estimate the maximum entropy autoregressive (AR) power spectrum of ultrasonic Doppler shift signals and display the results in the form of a sonogram in real-time on a computer screen. The system, which is based on a TMS 320C25 digital signal processor chip, calculates spectra with 128 frequency components from 64 samples of the Doppler signal. The samples are collected at a programmable rate of up to 40.96 kHz, and the computation of each spectrum takes typically 3.2 ms. The feasibility of on-line AR spectral estimation makes this type of analysis an attractive alternative to the more conventional fast Fourier transform approach to the analysis of Doppler ultrasound signals.     

Automatic detection of emboli in the TCD RF signal using principal component analysis

The transcranial Doppler (TCD) radio-frequency (RF) signal can provide additional information on events recorded during ultrasonic monitoring. Embolic signals appear as uniform and predictable shapes within the RF signal, enabling pattern recognition and image processing techniques to be used for their automated detection. This paper uses principal component analysis (PCA) to characterise the typical variation in embolic signal shape, within the RF signal, using training sets of in vitro and in vivo data. PCA techniques are then utilised to discriminate between previously unseen embolic and artifact signals. Although the results of this study show that the algorithms described in this paper do not yet have the accuracy required for their use in a clinical setting, it does demonstrate that this novel technique has the potential to be developed further. (E-mail: dhe@le.ac.uk)     

Customized spectral band analysis compared with conventional Fourier analysis of heart rate variability in neonates

A customized filtering technique is introduced and compared with fast Fourier transformation (FFT) for analyzing heart rate variability (HRV) in neonates from short-term recordings. FFT is classically the most commonly used spectral technique to investigate cardiovascular fluctuations. FFT requires stability of the physiological signal within a 300 s time window that is usually analyzed in adults. Preterm infants, however, show characteristics of rapidly fluctuating heart rate and blood pressure due to an immature autonomic regulation, resulting in non-stationarity of these signals. Therefore neonatal studies use (half-overlapping or moving) windows of 64 s length within a recording time of 2-5 min. The proposed filtering technique performs a filtering operation in the frequency range of interest before calculating the spectrum, which allows it to perform an analysis of shorter periods of only 42 s. The frequency bands of interest are 0.04-0.15 Hz (low frequency, LF) and 0.4-1.5 Hz (high frequency, HF). Although conventional FFT analysis as well as the proposed alternative technique result in errors in the estimation of LF power, due to spectral leakage from the very low frequencies, FFT analysis is more sensitive to this effect. The response times show comparable behavior for both the techniques. Applying both the methods to heart rate data obtained from a neonate before and after atropine administration (inducing a wide range of HRV), shows a very significant correlation between the two methods in estimating LF and HF power. We conclude that a customized filtering technique might be beneficial for analyzing HRV in neonates because it reduces the necessary time window for signal stability.     

Time-resolved pulsed elastography with ultrafast ultrasonic imaging

In this paper, a new elastographic method is proposed. Using this method, the propagation of a low-frequency transient shear wave can be imaged by means of an ultrafast imaging system (up to 10,000 frames/s) that we have developed. Ultrafast ultrasonic imaging is obtained with a linear array of transducers (3.5 MHz) connected to electronics that have 64 channels sampled at 30 MHz and 128 Kbytes for storing the backscattered signals. Displacements are measured using cross-correlation of the ultrasonic signals. Movies of the low central frequency (200 Hz) shear wave propagation through homogeneous and heterogeneous phantoms have been obtained with 1,000 and 2,000 frames per second.     

MP3 compression of Doppler ultrasound signals

The effect of lossy, MP3 compression on spectral parameters derived from Doppler ultrasound (US) signals was investigated. Compression was tested on signals acquired from two sources: 1. phase quadrature and 2. stereo audio directional output. A total of 11, 10-s acquisitions of Doppler US signal were collected from each source at three sites in a flow phantom. Doppler signals were digitized at 44.1 kHz and compressed using four grades of MP3 compression (in kilobits per second, kbps; compression ratios in brackets): 1400 kbps (uncompressed), 128 kbps (11:1), 64 kbps (22:1) and 32 kbps (44:1). Doppler spectra were characterized by peak velocity, mean velocity, spectral width, integrated power and ratio of spectral power between negative and positive velocities. The results suggest that MP3 compression on digital Doppler US signals is feasible at 128 kbps, with a resulting 11:1 compression ratio, without compromising clinically relevant information. Higher compression ratios led to significant differences for both signal sources when compared with the uncompressed signals.     

A new algorithm for off-line automated emboli detection based on the pseudo-wigner power distribution and the dual gate TCD technique

Research on microembolic signals (MES) using the dual-gate technique has shown promising results, when the time difference (Deltat) of a MES in two sample volumes (SVs) placed serially has been measured manually. On the other hand, the computerized discrimination of MES and artefacts has been reported not to be superior to algorithms based on a single SV. Therefore, a dataset containing MES as well as four types of artefacts was made to test a preliminary version of a new algorithm for automated emboli detection. We monitored 20 patients during carotid endarterectomy (n = 17) and heart surgery (n = 3). Two transcranial Doppler (TCD) signals with a partial overlap of the SVs were recorded online and analysed off-line with an algorithm based on three consecutive steps: 1. Is there an intensity increase in both channels (64-point FFT; 50% overlap)? 2. What is the expected time difference (Deltat), with the velocity measured in channel 1 as the calculation basis? 3. What is the 'exact' Deltat (pseudo-Wigner power function)? Two human experts decided whether a signal was a MES or belonged to one of the four artefact groups. Of a total of 97 MES, 28% (n = 27) could not be detected in the distal channel. Thus, 72% (n = 70) of the MES were present in both channels and could be analysed based on the abovementioned criteria. Of these 70 MES, 87% (n = 61) were correctly identified off-line. We assessed artefact rejection for four different types of artefacts: changes of TCD settings, probe movement, low flow artefacts and electrocautery. The reliability of artefact rejection was 98% for setting changes (n = 382), 96% for probe movement (n = 477) and 98% for low flow artefacts (n = 91), but only 68% for electrocautery (n = 264). These preliminary results are promising, but need careful interpretation: 28% of the MES were not detectable in the distal SV, probably due to a poor signal-to-noise ratio (SNR) and anatomical restrictions. Electrocautery signals were insufficiently rejected. However, even an artefact rejection of 96% can be insufficient if the number of MES is very small compared to the number of artefacts.     

Frequency filtering improves ultrasonic embolic signal detection.

Problems in detection of Doppler cerebral embolic signals primarily occur for embolic signals of low relative intensity. A characteristic feature of embolic signals is that the intensity increase is maximal over a narrow frequency band. Therefore, frequency filtering of the data might improve embolic signal relative intensity and detectability. We implemented an off-line finite impulse response filter in software running on a commercially available transcranial Doppler system, using the time-domain audio data as input. The range of the filter was chosen by placing a box around the embolic signal on the spectral display. One hundred consecutive embolic signals from patients with carotid stenosis were analyzed; all had been recorded by a bigate system and the signal was analyzed in both proximal and distal channels. There was a highly significant increase in embolic signal relative intensity following frequency filtering; mean (SD) proximal channel prefiltering 12.75 (4.83) dB, postfiltering 16.36 (4.93) dB; distal channel prefiltering 13.42 (4.98) dB, postfiltering 16.60 (5.11) dB, for both p < 0.001. Despite all embolic signals being audible and visible in at least one channel on the frequency spectral display, in 17 cases, the amplitude increase associated with the embolic signal could not be clearly seen in time-domain data of one or both channels prior to filtering. Following frequency filtering, this was reduced to 5. Incorporation of such a frequency-filtering approach to an online system is likely to improve the sensitivity of online detection for embolic signals of low relative intensity.     

Photoacoustic imaging with a commercial ultrasound system and a custom probe

Building photoacoustic imaging (PAI) systems by using stand-alone ultrasound (US) units makes it convenient to take advantage of the state-of-the-art ultrasonic technologies. However, the sometimes limited receiving sensitivity and the comparatively narrow bandwidth of commercial US probes may not be sufficient to acquire high quality photoacoustic images. In this work, a high-speed PAI system has been developed using a commercial US unit and a custom built 128-element piezoelectric-polymer array (PPA) probe using a P(VDF-TrFE) film and flexible circuit to define the elements. Since the US unit supports simultaneous signal acquisition from 64 parallel receive channels, PAI data for synthetic image formation from a 64- or 128-element array aperture can be acquired after a single or dual laser firing, respectively. Therefore, two-dimensional (2-D) B-scan imaging can be achieved with a maximum frame rate up to 10 Hz, limited only by the laser repetition rate. The uniquely properties of P(VDF-TrFE) facilitated a wide -6 dB receiving bandwidth of over 120% for the array. A specially designed 128-channel preamplifier board made the connection between the array and the system cable, which not only enabled element electrical impedance matching but also further elevated the signal-to-noise ratio (SNR) to further enhance the detection of weak photoacoustic signals. Through the experiments on phantoms and rabbit ears, the good performance of this PAI system was demonstrated. (E-mail: xdwang@umich.edu)     

Detection of Doppler embolic signals: psychoacoustic considerations

The purpose of this study was to improve reliability in the identification of Doppler embolic signals by determining the decibel threshold for reproducible detection of simulated "emboli" as a function of signal duration, frequency and cardiac-cycle position. The auditory sensitivity of 16 participants to 574 simulated "emboli" was examined using psychoacoustic techniques to assess how the probability of detection varies with embolic signal parameters. Detailed measurements of the threshold for detection of simulated embolic signals are presented. These provide evidence that the measured embolus-to-blood threshold ranges between 2 dB and 14 dB as a continuous function of signal duration and frequency. The level of the threshold is closely linked to both embolic signal parameters and the properties of the blood flow signal. We conclude that the current fixed choice of threshold does not provide a good approximation to the true threshold of detection across the full range of embolic signal parameters.     

高场强磁共振系统点分辨波谱技术探讨

目的研究单体素点分辨波谱技术 (PRESS) 中不同参数对结果的影响,并对优化技术参数进行探讨.方法运用GE 3.0 T signa excite 核磁共振扫描仪对四组无明显临床脑神经系统疾患者分别行单体素PRESS扫描.每组在其他参数不变的条件下,分别改变不同的TR, NAV, NEX以及在静注造影剂前后扫描.记录NAA/Cr, Cho/Cr, mI/Cr和SNR值,并进行统计学分析.如数据属正态分布,则进行配对t检验;否则进行Wilcoxon检验.结果各组中NAA/Cr, Cho/Cr和mI/Cr均没有显著性差异.TR为 2000 ms时,信噪比(SNR)明显高于TR 为1500 ms和1200 ms时.NAV为128时,SNR明显高于NAV为64和32时.但NAV为128和96,NEX为 8和2以及静注造影剂前后SNR没有显著性差异.结论在PRESS序列参数的选择上,考虑到SNR、测量值及扫描时间,最佳参数为TR 1500 ms,TE 35 ms,NAV 96和NEX 2.在某些特定情况下,参数可做相应调整.     

Subsample volume processing of doppler ultrasound signals

Processing of Doppler signals produced by pulsed Doppler systems is based on the assumption that the phase of the received high frequency ultrasound signals changes linearly with depth. However, the random spatial distribution of scatterers is not in accordance with this basic assumption. Consequently, averaging of the demodulated signal over an observation window, covering a few periods of the received signal, does not improve the estimate for the instantaneous quadrature components of the Doppler signal originating from a given depth. Hence, the accuracy of the Doppler velocity estimate is independent of the length of the observation window employed. However, splitting the observation window in subsample volumes, each with a length of one period at the emission frequency, and combining the Doppler signals of the subsample volumes at the last stage of signal processing, i.e., mean Doppler frequency estimation using the autocorrelation technique, results in a considerable reduction of the variance of the velocity estimate. Using a computer simulation of the signal processing involved, it is demonstrated that with subsample volume processing the variance of the velocity estimate attains the same variance as is expected for the RF cross correlation technique.     

Determination of tissue motion velocity by correlation interpolation of pulsed ultrasonic echo signals

Correlation interpolation is introduced as a method to determine the displacement of moving biological tissue on the basis of a sequence of ultrasonic echo signals. The echo signal is sampled along the echo depth with approximately 4 samples per average high frequency period. Sampling in time occurs with the pulse repetition frequency. The necessary information is extracted from a crosscorrelation function between successive signals, which is modelled using four parameters. The parameters are estimated from five calculated correlation sums and the shift with maximum correlation is determined. In contrast to existing techniques, the performance of this method is determined mainly by the number of samples used, while the ratio of the number of samples in depth and time is irrelevant. Using 64 samples at a signal-to-noise power ratio of 10, the standard deviation of the error in the determination of the shift in depth is 0.08 sampling intervals. As in many other methods, the width of the aliasing interval equals the mean frequency period.     

A comparative study and assessment of Doppler ultrasound spectral estimation techniques. Part I: Estimation methods

When compared to the classical Discrete Fourier Transform (DFT) or Fast Fourier Transform (FFT) approach, modern estimation methods offer the potential for achieving significant improvements in estimating the power density spectrum of Doppler ultrasound signals. Such improvements, for example, might enable minor flow disturbances to be detected, thereby improving the sensitivity in arterial disease assessment. Specifically, reduction in the variance and bias can be achieved, and this may enable disturbed flow to be detected in a more sensitive manner. The approach taken here, is to consider spectral estimation methods as a problem of fitting an assumed model to the Doppler signal. The models described assume that the signal is stationary. Since the Doppler signal is generally nonstationary, it is assumed that a short enough time window interval can be chosen over which the signal can be considered stationary. We shall review the various methods and when appropriate, relate them to the nature of the Doppler signal.     

Characteristics of Doppler embolic signals observed following carotid endarterectomy

Postoperative Doppler embolic signals following carotid endarterectomy (CEA) are associated with an increased risk of stroke, but the characteristics of these signals are rarely reported. In this study, we survey signals from 1485 emboli, assumed to consist predominantly of thrombus. Data were obtained by monitoring the middle cerebral arteries of 100 consecutive CEA patients during postoperative recovery. The distribution of embolic signal frequencies, intensities and durations revealed that embolic signals do not occur randomly in the sonogram. In particular, we find that the signals possess a characteristic distribution of velocities reflecting the preferred path of the embolus through the artery (at approximately 75% of the distance between the centre of the artery and the artery wall). Embolic signals were more likely to be observed at cardiac cycle positions between 35% and 80% from the start of systole than elsewhere. After eliminating other considerations, we hypothesized that this peak in the distribution of signals in the sonogram arose due to the localization of emboli trajectories and a strong tendency for emboli to detach from the carotid bifurcation during systole.     

Spectral domain optical coherence tomography of multi-MHz A-scan rates at 1310 nm range and real-time 4D-display up to 41 volumes/second

An ultrafast frequency domain optical coherence tomography system was developed at A-scan rates between 2.5 and 10 MHz, a B-scan rate of 4 or 8 kHz, and volume-rates between 12 and 41 volumes/second. In the case of the worst duty ratio of 10%, the averaged A-scan rate was 1 MHz. Two optical demultiplexers at a center wavelength of 1310 nm were used for linear-k spectral dispersion and simultaneous differential signal detection at 320 wavelengths. The depth-range, sensitivity, sensitivity roll-off by 6 dB, and axial resolution were 4 mm, 97 dB, 6 mm, and 23 μm, respectively. Using FPGAs for FFT and a GPU for volume rendering, a real-time 4D display was demonstrated at a rate up to 41 volumes/second for an image size of 256 (axial) × 128 × 128 (lateral) voxels.OCIS codes: (100.6890) Three-dimensional image processing, (110.4500) Optical coherence tomography, (170.4500) Optical coherence tomography     

RF signals provide additional information on embolic events recorded during TCD monitoring

Transcranial Doppler ultrasound (TCD) is commonly used to detect embolic signals in the cerebral circulation. However, current techniques to discriminate between signals from emboli and artifacts are subjective and ambiguous. The radiofrequency (RF) signal provides an extra dimension to the information available from conventional TCD systems that may help to interpret complex events. Artifacts generated by healthy volunteers and embolic signals recorded from a flow phantom were used to characterize the appearance of the two types of event. Characteristics of events, recorded during and immediately after carotid endarterectomy surgery, were compared with those from known sources. Additional information was provided by the RF signal on events recorded during TCD monitoring thus aiding classification. The RF signal may have a role as a "gold standard" for embolus detection.     

A dynamic system model-based technique for functional MRI data analysis

Signals in functional magnetic resonance imaging (fMRI) are influenced by physiological fluctuations in addition to local brain activity. We have proposed a dynamic system model-based technique for separation of signal changes related to brain activation inputs from those related to physiological fluctuations. We applied this technique to a visual fMRI experiment to determine the validity and feasibility of this technique for fMRI data analyses. Gradient-echo echo planar images were obtained from 12 healthy volunteers with a Siemens ALLEGRA operating at 3 T, with a repetition time of 500 ms, echo time of 20 ms, field of view of 200-210 mm, matrix size of 64 x 64, and slice thickness of 5 mm. Twelve runs with two stimulation periods of varied duration (2-8 s) with 8-Hz flickering illumination were obtained for each subject. Local signal changes were modeled by an autoregressive model with two exogenous inputs, a visual stimulation input and a global reference signal. Local signal changes were appropriately predicted not only for stimulation periods but also resting periods. A significant linear relationship was found between model static gain based on the dynamic system modeling and beta coefficient based on a general linear model (GLM) analysis for active voxels in the primary visual cortex (analysis of covariance [ANCOVA], P < 0.001; estimated parameter, 0.967; 95% confidence interval, 0.734-1.201). This dynamic system model-based technique is sufficiently accurate and feasible for use in extracting signal changes related to brain activation inputs from measured signals with physiological fluctuations.     

Myocardial rapid velocity distribution

Myocardial motion exhibits frequency components of up to 100 Hz, as found by a phased tracking method. To simultaneously measure the rapid and minute velocity signals at multiple points along the surface of the left ventricle (LV), in this study, conventional ultrasonic diagnosis equipment was modified to allow 10 scan lines from a sector scanner to be arbitrarily selected in real-time for analysis. By considering the maximum value of the velocity in the heart wall and the maximum depth from the chest surface, the number of transmission directions of the ultrasonic pulses should be carefully confirmed to be 10 to avoid aliasing, which is much less than the number employed in conventional tissue Doppler imaging (TDI). By applying the system, the velocity signals at about 240 points in the heart walls were simultaneously measured for three healthy volunteers. During a short period of 35 ms around end-diastole, the velocity signals varied spatially in the heart wall. At the end of systole, in the wavelets near the base of the interventricular septum (IVS), the slow pulse continued for about 30 ms, just before the radiation timing of the second heart sound. Then, a steep pulse occurred just at the timing of the closure of the aortic valve. The steep pulse at the base preceded that at the apex by several ms. By Fourier transforming each wavelet, the spatial distribution of the phase of the steep pulse components were clearly displayed. By applying the measurement method to two patients with aortic stenosis (AS), irregular vibration signals, which correspond to the murmur of the heart sound, could be directly detected during the ejection period. In conventional TDI, only the large slow movements due to the heartbeat are displayed, but these rapid and minute velocity components cannot be displayed. In this study, moreover, the phase components were detected for the first time from each of the velocity signals simultaneously measured at multiple points along the 10 scan lines. This measurement and method of analysis offer potential for new diagnostic techniques in cardiac dysfunction.     

Nonstationarity broadening in pulsed Doppler spectrum measurements

Conventional measurement of the spectrum of arterial signals from the pulsed ultrasonic Doppler instrument uses windowed, sequential data segments. The Doppler signal is assumed stationary for the duration of each segment. It is shown here that this assumption is often unreasonable and the effect of mean frequency variation during the data segment has been investigated for different windows and rates of change of mean frequency. A data segment length giving maximum spectral resolution is shown to exist for each window type and rate of frequency change.