Robert W. Gill

An axial velocity estimator for ultrasound blood flow imaging, based on a full evaluation of the Doppler equation by means of a two-dimensional autocorrelation approach

This paper introduces a new velocity estimator, referred to as the 2D autocorrelator, which differs from conventional Doppler techniques in two respects: the derivation of axial velocity values by evaluating the Doppler equation using explicit estimates of both the mean Doppler and the mean RF frequency at each range gate location; and, the 2D nature (depth samples versus pulse transmissions) of processing within the range gate. The estimators output can be calculated by evaluating the 2D autocorrelation function of the demodulated (baseband) backscattered echoes at two lags. A full derivation and mathematical description of the estimator is presented, based on the framework of the 2D Fourier transform. The same framework is adopted to analyze two other established velocity estimators (the conventional 1D autocorrelator and the crosscorrelator) in a unifying manner, and theoretical arguments as well as experimental results are used to highlight the common aspects of all three estimators. In addition, a thorough performance evaluation is carried out by means of extensive simulations, which document the effect of a number of factors (velocity spread, range gate length, ensemble length, noise level, transmitted bandwidth) and provide an insight into the optimum parameters and trade-offs associated with individual algorithms. Overall, the 2D autocorrelator is shown to offer the best performance in the context of the specific simulation conditions considered here. Its superiority over the crosscorrelator is restricted to cases of low signal-to-noise ratios. However, the 2D autocorrelator always outperforms the conventional 1D autocorrelator by a significant margin. These comparisons, when linked to the computational requirements of the proposed estimator, suggest that it combines the generally higher performance of 2D broadband time-domain techniques with the relatively modest complexity of 1D narrowband phase-domain velocity estimators.<<ETX>>

Experimental evaluation of velocity and power estimation for ultrasound blood flow imaging, by means of a two-dimensional autocorrelation approach

This paper evaluates experimentally the performance of a novel axial velocity estimator, the 2D autocorrelator, and its Doppler power estimation counterpart, the 2D zero-lag autocorrelator, in the context of ultrasound color flow mapping. The evaluation also encompasses the well-established 1D autocorrelation technique for velocity estimation and its corresponding power estimator (1D zero-lag autocorrelator), to allow performance comparisons under identical conditions. Clutter-suppressed in vitro data sets from a steady-flow system are used to document the effect of the range gate and ensemble length, noise level and angle of insonation on the precision of the velocity estimates. The same data sets are used to examine issues related to the estimation of the Doppler signals power. The first-order statistics of power estimates from regions corresponding to flow and noise are determined experimentally and the ability of power-based thresholding to separate flow signals from noise is characterized by means of ROC analysis. In summary, the results of the in vitro evaluation show that the proposed 2D-autocorrelation form of processing is consistently better than the corresponding 1D-autocorrelation techniques, in terms of both velocity and power estimation. Therefore, given their relatively modest implementation requirements, the 2D-autocorrelation algorithms for velocity and power estimation appear to represent a superior, yet realistic, alternative to conventional Doppler processing for color flow mapping.<<ETX>>

Knowledge-based method for segmentation and analysis of lung boundaries in chest X-ray images.

We present a knowledge-based approach to segmentation and analysis of the lung boundaries in chest X-rays. Image edges are matched to an anatomical model of the lung boundary using parametric features. A modular system architecture was developed which incorporates the model, image processing routines, an inference engine and a blackboard. Edges associated with the lung boundary are automatically identified and abnormal features are reported. In preliminary testing on 14 images for a set of 18 detectable abnormalities, the system showed a sensitivity of 88% and a specificity of 95% when compared with assessment by an experienced radiologist.

Transvaginal pulsed Doppler ultrasound assessment of blood flow to the corpus luteum in IVF patients following embryo transfer

Summary. The waveforms of vessels supplying the ovaries of women on an in-vitro fertilization (IVF) programme were studied using transvaginal B-mode and Doppler ultrasound. There were 125 scans recorded in 65 women at weekly intervals from 3 days after embryo transfer or 5 days after gamete intrafallopian transfer (GIFT) until confirmation of pregnancy or onset of menses. At each examination the signals obtained from vessels supplying the ovaries were recorded and quantified using a resistance index (RI). Fifteen patients became pregnant of whom one has subsequently miscarried. There was a highly significant difference in the RI values between patients who became pregnant and those who did not; no patient who became pregnant had a RI greater than 0.5. Oestrogen to progesterone ratios were calculated in the subgroup of non-pregnant patients and there was no correlation between these values and the RI values. This new technique enables prediction of IVF treatment failure earlier than has been reported previously and may reflect the inadequacy of the corpus luteum.

Model-based assessment of lung structures: inferencing and control system

A general methodology has been developed for computer interpretation of medical images, based on an explicit anatomical model. A test system for analyzing posterior- anterior (PA) chest x-rays has been implemented. The inferencing and control system identifies the major lung structures in the image, and then flags any suspected abnormalities. Image and model data are transformed into a feature space where they are represented in terms of edge descriptions. The inference engine compares the image and model in feature space to label the edges anatomically, and check for normality. The control system schedules events within the inference engine and coordinates interaction with the model and image processing routines. The control architecture is blackboard-based, with a separate data frame for each structure to be identified. The anatomical model uses fuzzy sets to provide ranges of feature values which are considered normal or indicative of a particular abnormality. This allows the inference engine to give a confidence score and linguistic description to each decision. Mediastinum, cardiac border, domes of the diaphragm, ribs and lung outline have been modeled. Their automatic identification allows diagnostic checks such as the cardiothoracic ratio, comparison of right and left lungs to identify lobular collapse and inspection of interfaces in terms of shape and clarity. The inference engine provides simple comments on its findings, making it suitable for pre- and double-checking of images.

Automatic vessel tracking and measurement for Doppler studies.

In this initial report, an imaging technique is described for automatically recognising the walls of a blood vessel during Doppler measurements, and hence automatically adjusting the position of the sample volume. The method reduces the inherently two-dimensional edge detection problem to a one-dimensional one, and iterates over several imaging frames to optimise the information. The orientation and diameter of the vessel are also measured. The technique has been clinically tested off-line and is particularly applicable to Doppler volumetric flow studies.

Computer analysis of the early diastolic notch in doppler sonograms of the uterine arteries

This article describes a set of processing and analysis techniques for automated identification and quantification of the early diastolic notch (EDN), a feature of Doppler sonograms from the uterine arteries which has been associated with adverse pregnancy outcomes such as preeclampsia and intrauterine growth retardation. Examples covering different sonogram types are provided to illustrate the effectiveness and reproducibility of the processing/analysis tools. Also, a receiver-operating characteristic-based evaluation of the EDN quantification and pulsatility indexes is presented, which examines the ability to predict hypertension and/or intrauterine growth retardation, using a set of uterine Doppler sonograms from 92 patients acquired at 18 weeks of gestation. In summary, the ROC results confirm the link between the EDN and abnormal pregnancy outcomes, and suggest that EDN quantification has a higher diagnostic accuracy than the pulsatility index, which characterises the flow waveform in a global manner and therefore does not take explicitly into account the localised nature of the EDN. Quantification of the EDN at 18 weeks of gestation appears to best predict the most severely abnormal pregnancy outcomes.

A comparison of matched signals used in three different phase-aberration correction algorithms

The Nearest-Neighbor Cross-Correlation (NNCC), Translating Apertures (TA), and Near-Field Signal Redundancy (NFSR) algorithms have a common feature: they all calculate cross-correlation functions between signals that are assumed to be highly correlated (matched signals), and then derive the aberration profile from the peak positions of these cross-correlation functions. One of the major differences between them is the way matched signals are collected. In this paper, a sub-signal analysis of matched signals is performed to demonstrate the different sub-signal components in matched signals collected with these three algorithms. The sub-signal components of matched signals influence the similarity between them and they have significant impact on the performance of an algorithm. The similarity between matched signals collected with these three algorithms is experimentally compared by calculating the cross-correlation coefficient between matched signals. Signals are collected from a phantom with a modified ATL Ultramark(R) 8 ultrasound scanner. It shows that the degree of similarity between matched signals in these algorithms is in the following order (from more to less similar): NFSR, TA, and NNCC. It also shows that, when phase aberrations exist, the cross-correlation coefficients between matched signals in the NNCC and TA algorithms decrease more dramatically than those in the NFSR algorithm.

Correction of distributed phase aberrations caused by propagation-speed and array-pitch errors

When an incorrectly assumed array pitch or propagation speed is used to form an image, phase aberrations are generated even in a homogeneous medium, and they are distributed. That is, at different image points the phase aberration profiles are different. Even though array pitch and propagation speed are usually known quite accurately, it is important and interesting to test whether a phase aberration correction algorithm is capable of correcting distributed phase aberrations caused by incorrect array pitch and propagation speed values. It was demonstrated that the Near Field Signal Redundancy (NFSR) algorithm is able to correct these kinds of phase aberrations with data from a linear array. However, only a single phase-aberration profile was measured and the image quality after correction was only improved in a small region. In this paper we present new results from a phased array where multiple phase aberration profiles have been measured and the whole image has been corrected. These results have further demonstrated that the NFSR algorithm is capable of correcting distributed phase aberrations caused by array pitch and propagation speed errors.