Freehand 3D ultrasound is particularly appropriate for the measurement of organ volumes. For small organs, which can be fully examined with a single sweep of the ultrasound probe, the results are known to be much more accurate than those using conventional 2D ultrasound. However, large or complex shaped organs are difficult to quantify in this manner because multiple sweeps are required to cover the entire organ. Typically, there are significant registration errors between the various sweeps, which generate artifacts in an interpolated voxel array, making segmentation of the organ very difficult. This paper describes how sequential freehand 3D ultrasound can be used to measure the volume of large organs without the need for an interpolated voxel array. The method is robust to registration errors and sweep orientation, as demonstrated in simulation and also using in vivo scans of a human liver, where a volume measurement precision of plus or minus 5% is achieved.

The user interface is critical to the clinical acceptance of free-hand 3D ultrasound imaging. In the system described here, the acquisition of scans from a conventional 2D probe is synchronized to probe movement. Motion-gated acquisition reduces the dependency of scan quality on the operator and can detect poor scanning technique. Automatic annotation is facilitated by recording the orientation of the patient. A 'virtual probe' enables a clinician to reslice the acquired data as if the patient were being rescanned. A 'thick reslice' enables spatial compounding to be implemented in the visualization process. The critical scale parameter, which is difficult to automatically determine, is intuitively interpreted as the slice thickness and interactively controlled by the clinician. Real-time volume rendering without prior segmentation of the data is possible using this technique. Shape-based interpolation assists manual segmentation of the data. The distance field from the segmentation process is used to mask the raw data and the result is volume rendered. A 3D scalpel enables efficient refinement of the segmentation by directly editing the volume- rendered view. The 3D scalpel prunes a set of user-defined contours to which a surface is fitted and the enclosed volume computed. A compact set of contours is retained by the system which allows the actions of the user to be undone. Three clinicians have tested the system, acquiring over 100 3D fetal scans, in a hospital environment.

Visualization of volumetric medical data is challenging. Rapid-prototyping (RP) equipment producing solid object prototype models of computer generated structures is directly applicable to visualization of medical anatomic data. The purpose of this study was to develop methods for transferring 3D Ultrasound (3DUS) data to RP equipment for visualization of patient anatomy. 3DUS data were acquired using research and clinical scanning systems. Scaling information was preserved and the data were segmented using threshold and local operators to extract features of interest, converted from voxel raster coordinate format to a set of polygons representing an iso-surface and transferred to the RP machine to create a solid 3D object. Fabrication required 30 to 60 minutes depending on object size and complexity. After creation the model could be touched and viewed. A '3D visualization hardcopy device' has advantages for conveying spatial relations compared to visualization using computer display systems. The hardcopy model may be used for teaching or therapy planning. Objects may be produced at the exact dimension of the original object or scaled up (or down) to facilitate matching the viewers reference frame more optimally. RP models represent a useful means of communicating important information in a tangible fashion to patients and physicians.

Very high frequency (50 MHz) ultrasound provides spatial resolution on the order of 30 microns axially by 60 microns laterally. Our aim was to reconstruct the three-dimensional anatomy of the eye in the full detail permitted by this fine- scale transducer resolution. We scanned the eyes of human subjects and anesthetized rabbits in a sequence of parallel planes 50 microns apart. Within each scan plane, vectors were also spaced 50 microns apart. Radio-frequency data were digitized at a rate of 250 MHz or higher. A series of spectrum analysis and segmentation algorithms was applied to data acquired in each plane; the outputs of these procedures were used to produce color-coded 3-D representations of the sclera, iris and ciliary processes to enhance 3-D volume rendered presentation. We visualized the radial pattern of individual ciliary processes in humans and rabbits and the geodetic web of supporting connections between the ciliary processes and iris that exist only in the rabbit. By acquiring data such that adjacent vectors and planes are separated by less than the transducer's lateral resolution, we were able to visualize structures, such as the ciliary web, that had not been seen before in-vivo. Our techniques offer the possibility of high- precision imaging and measurement of anterior segment structures. This would be relevant in monitoring of glaucoma, tumors, foreign bodies and other clinical conditions.

Dynamic-3D echocardiography can potentially provide an accurate representation of the complex cardiac anatomy and dynamics over the cardiac cycle. However, the image quality of dynamic-3D echocardiography is limited by temporal jitter artifacts; which result from the asynchronous acquisition of video frames with respect to the cardiac cycle. In our study, we estimate the extent of temporal jitter in dynamic-3D echocardiography using in-vitro studies, and also provide a theoretical method to predict temporal jitter. Dynamic-3D images of a myocardial motion phantom were reconstructed and analyzed for cardiac wall motion at different heart rates. Our algorithm to quantify temporal jitter consisted of three steps. First, the distance of a reference plane to the surface of the phantom was computed and plotted as a 2D grayscale distance map in each 3D image. Second, surface variation maps were derived and finally, 2D jitter maps were plotted to provide a measure of temporal jitter in each 3D image at each cardiac phase. Temporal jitter was as large as 3.5 mm in peak systole and 1.5 mm in end diastole. The standard deviation in the jitter maps ranged from 0.5 mm in end-diastole to 1.0 mm in peak-systole; which agreed well with our theoretical analysis.

A technique is introduced which determines the spacing between acquired 2D planes necessary to maintain resolution in the scan direction while maximizing scan speed for 3D ultrasound imaging. A significant limit on resolution in ultrasound systems is the granular interference pattern called speckle. We perform a statistical analysis of the speckle in a series of 3D scans of agar blocks with different image spacings. Speckle size was approximately constant at small 2D image- plane spacings, but increased once the scan spacing surpassed a critical point. This point is the optimum scan spacing, minimizing the detrimental effects of speckle, while maximizing scan speed. Optimum spacing values range from .075 to .4 mm increasing with the axial depth and the number of focal zones. This dependence on the number of focal zones is a result of successive sampling by the US machine and the digitizer at unequal rates, as demonstrated through an analysis of the noise power spectra. Such sampling is commonly used in 3D ultrasound and inflates the speckle size. Our analysis predicted that speckle size may be significantly reduced by using a transducer sampling rate that is twice the digitization rate.

Two- and three-dimensional depictions of ultrasound echo signal data have potential for helping to detect and diagnose disease and to plan and monitor therapy. The utilization of very-high-frequency ultrasound and spectrum analysis of radio- frequency echo signals extends the capabilities of ultrasonic imaging for these purposes. Images generated using these techniques can present tissue architecture with exquisite resolution and can provide information on underlying properties of scatterers in the tissue. Changes in properties over time can be used to monitor disease progression or response to therapy. Relating tissue echo-signal parameters obtained from unknown tissue to database values of known tissue types can provide means of characterizing tissue for the purposes of detection or diagnosis and treatment planning. These potential applications are illustrated using examples from plaque, ophthalmic, skin, and prostate studies.

The usefulness of stereotactic neurosurgery performed via a craniotomy is limited because the craniotomy leads to a brain tissue shift of 10 mm on average. We have recently completed an examination of 2D intra-operative ultrasound as a means of visualization and measurement of brain shift. A commercial 3D tracking system was used for real-time registration of the ultrasound video to pre-operative MR images, and annotation of the images was used to measure the shift. More than 15 surgical cases have been performed thus far with the 2D system. We are now undertaking phantom studies with tracked 3D ultrasound, and have developed sophisticated tools for real- time overlay of ultrasound and MRI volumes. These tools include a virtual-reality view of the ultrasound probe with live ultrasound video superimposed over a 3D -rendered MRI of the brain, as well as 3D ultrasound/MRI transparency overlay views. Algorithms to automatically extract landmarks from MRI and 3D ultrasound images are under development. We aim to use these landmarks to automatically generate nonlinear warp transformations to correct the pre-operative MRI as well as surgical target coordinates for brain shift. Portions of the C++ code developed for this project have been contributed to the open-source Visualization Toolkit (VTK).

The design and fabrication of high frequency single element ultrasonic transducers present a multitude of challenges for the transducer engineer, from size constraints to electrical impedance matching. This paper discusses the trade-offs involved in a procedure used to fabricate transducers with center frequencies in the 25 MHz - 100 MHz range using 36 degree rotated Y-cut lithium niobate (LiNbO₃) as the active element. Transducers of varying dimensions were built according to an f-number range of 2 - 3.5. A (lambda) /4 silver epoxy matching layer with an acoustic impedance of 7.3 Mrayls was used. Desired focal depths were achieved with use of an acoustic lens. Silver epoxy backing with an acoustic impedance of approximately 5.9 Mrayls was also used in all designs. All transducers were designed around a 50(Omega) send and receive circuit. Electrical tuning of the transducer to the receive circuitry was achieved by using an RF transformer. All transducers were tested in a pulse-echo arrangement using a Panametrics 5900PR pulser, a Wavetek function generator and a LeCroy digital oscilloscope. The bandwidth, insertion loss, and depth of focus were measured. Several transducers were fabricated with -6dB bandwidths ranging from 62% to 74%, and two-way insertion loss values ranging from -14dB to -22dB.

A wide variety of fabrication techniques and materials produce ultrasound transducers with very different performance characteristics. High frequency (50 MHz), focused single element transducers using lead zirconate titanate (PZT) fiber composite, lead titanate (PbTiO₃) ceramic, polyvinylidene fluoride (PVDF) polymer and lithium niobate (LiNbO₃) single crystal are compared in design and performance. The transducers were all constructed with a 3 mm aperture and f- number of 2 - 3. Design considerations discussed include optimization of designs using different lens, backing and matching materials for acoustic matching and the use of several electrical tuning techniques to match the transducers to the 50(Omega) circuitry. Transducers were tested for insertion loss and -6dB bandwidth using a quartz flat- plate target. Insertion loss measurements between transducers were -20dB to -50dB with bandwidths in the range of 50 - 120%. Through the use of an ultrasound backscatter microscope (UBM), the transducer were compared using in vitro images of the human eye. Images of a wire phantom were also made for comparison of lateral and axial resolution of each device.

Fine scale 1 - 3 fiber composites and PbTiO3 are promising materials for high frequency (greater than 30 MHz) ultrasonic transducers because of their high thickness coupling coefficients and good mechanical flexibility. This work focus on improvements in the fiber composite and PbTiO₃ ceramic transducer performance through innovative tuning techniques and optimized materials selection. For the fiber composite transducers, a 45% volume fraction of PZT fibers with 17 micrometer fiber diameter was selected to obtain high coupling (approximately 0.60) and ensure pure thickness vibration mode at high frequency. Microballoons mixed with epoxy formed a low acoustic impedance backing material and parylene was deposited as a front matching layer. The PbTiO₃ ceramic was first processed to provide high density and then polished to the required thickness. Conductive epoxy was then chosen as backing and parylene was deposited as a front matching layer. The focus for both transducers was achieved by spherical shaping. The fiber composite transducers with center frequency around 40 MHz showed 6 dB bandwidth as high as 100%, and round-trip insertion losses near -36 dB. Compared to fiber composite transducers, the PbTiO₃ ceramic transducers can work at much higher frequencies (near 90 MHz). In a similar frequency range as fiber composite transducers, they showed improvement sensitivity as marked by an insertion loss around -24 dB. In addition, they still displayed good bandwidth as a result of electrical tuning.

A new method for computing images of transient ultrasound fields from transducers of arbitrary shape is developed. For simplicity, only transducers with axial symmetry are considered here, but the extension to square and rectangular radiators is straightforward. The more general case may be treated by the same methods. The method is based on the use of the directivity spectrum -- which can be shown to be a generalization of the angular spectrum. It is ideally suited, however, for application to transient fields, and the formalism contains no evanescent waves. Images of pulses over extensive ranges from a variety of transducers are shown. In particular, it is shown that the transient field from a strongly-focused bowl transducer may be readily calculated, without the approximations that are necessary when using the traditional Tupholme-Stepanishen method. The simulation method is powerful and computationally efficient. It is considerably faster than methods used up to now, and may be applied to the computation of fields that are problematic for standard methods. The final output shown here is a high-quality 'snapshot' of the field, at various distances from the transducer face. Phase of the field is shown.

Polyvinylidene fluoride (PVDF) transducers have been developed for ultrasonic backscatter microscopy (UBM) applications. Single element devices encompassed the frequency range of 40 to 65 MHz and incorporated a variety of backing materials and tuning circuits. Pulse-echo testing was performed using the 1 (mu) J energy setting on a Panametrics 5900 pulser/receiver, configured with +40dB gain and 20 dB attenuation. A broad spectrum of performance was observed depending on the choice of backing and tuning, with -6dB bandwidths of 55% to 116% and echo amplitudes from 100mVpp to 940mVpp. It was observed that electrical impedance matching increased sensitivity and tuned the center frequency at the expense of reduced bandwidth. For imaging applications the tuned devices provided superior performance. Although single element transducers in this frequency range provide axial resolution on the order of 50 micrometer, a tradeoff exists between the lateral resolution and the depth of field. To obtain high lateral resolution and a long depth of field an annular array design was adopted. Laser dicing was used to fabricate the elements of the array from a sheet of PVDF. Interconnect to each element was achieved using pin contacts molded into the backing. The design incorporated a 5 mm aperture with six equal area elements focused at 10 mm. At a center frequency of 50 MHz this design will achieve a lateral resolution of 58 micrometer, comparable to the axial resolution, over the entire 7.5 mm depth of field.

Spectrum analysis of ultrasonic radio-frequency echo signals has proven to be an effective means of characterizing tissues of the eye and liver, thrombi, plaque, etc. Such characterization can be of value in detecting, differentiating, and monitoring disease. In some clinical applications, linear methods of tissue classification cannot adequately differentiate among the various manifestations of cancerous and non-cancerous tissue; in these cases, non-linear methods, such as neural-networks, are required for tissue typing. Combining spectrum-analysis methods for quantitatively characterizing tissue properties with neural-network methods for classifying tissue, a powerful new means of guiding biopsies, targeting therapy, and monitoring treatment may be available. Current studies are investigating potential applications of these methods that use novel tissue-typing images presented in two and three dimensions. Results to date show significant sensitivity improvements of possible benefit in cancer detection and effective tissue-type imaging that promise improved means of planning and monitoring treatment of prostate cancer.

We are combining techniques of quantitative ultrasonic imaging to study polycystic kidney disease (PKD) as the disease progresses to renal failure. Our goal is to use ultrasound noninvasively to detect morphological changes early in the disease process when interventions are most likely to be successful and prior to a significant loss in renal function. We are examining the kidneys of normal rats and those with PKD at various ages with several techniques to obtain comprehensive knowledge of the disease progression. The Han:SPRD rat inherits PKD as an autosomal dominant trait (ADPKD) that closely mimics ADPKD in humans. Changes in renal function are assessed using tracer kinetics (DTPA) and IOH clearance). Ultrasonic techniques, based on measurements of acoustic backscatter coefficients and parameters derived from these measurements, are sensitive to microscopic changes in the tissue morphology. Elasticity imaging is used to study the changes in the tissue macrostructure. All acoustic measurements are made using a state-of-the-art clinical imaging system (Siemens Elegra). Our results show that ultrasonic techniques are very sensitive to early changes in renal microstructure and macrostructure. Ultrasound can be used to detect changes in the renal cortex long before there is a measurable loss of renal function. These techniques are also useful for monitoring the progression of the disease. Most importantly, these techniques are noninvasive and directly applicable to humans.

Described are quantitative results of an ultrasound imaging method for discrimination between breast cancers and benign lesions. The procedure, called disparity mapping, may provide better medicine at lower cost. 27 in vivo samples were obtained from the Radiology Dept., Hospital of the University of Pennsylvania, of which 12 were cancers or were suspicious of being cancers and 15 were benign. Zero errors resulted from the procedure described herein. Undue optimism is unwarranted because of the small sample size, particularly of the cancers, and because the test was not blind. Because DM appears to react to elastic surface characteristics of lesions it also has the potential to disclose sites of active growth on cancerous lesions. This information, prior to surgery, would be valuable to the surgeon in planning the procedure.

The composition and morphology of the atherosclerotic lesion are considered to be important determinants of acute coronary ischemic syndromes. We investigated the potentials of a combination of intravascular ultrasound (IVUS) elastography and intravascular Near Infrared Raman (NIR) spectroscopy, to assess the physical and chemical composition of the vessel wall and plaque. Intact human coronary arteries were mounted in an in vitro pressurized perfusion setup and investigated with a 20 MHz Visions^R IVUS catheter. At selected cross- sections, two echo-frames were acquired at intraluminal pressures of 80 and 100 mmHg to strain the tissue in order to obtain elastograms. Next, Raman spectra were obtained during 30 seconds at 4 angles (0, 90, 180 and 270 degrees) using a sideways viewing probe. Spectra were modeled to obtain quantitative chemical information, while leaving the specimens intact. Calcified areas were identifiable on the echograms, elastograms and Raman spectra. A combination of geometric information provided by the echogram, chemical information as obtained with Raman spectroscopy, and high stress regions determined by the elastogram, may prove to be a valuable tool to identify plaque vulnerability.

Transit-time-based methods for ultrasonic velocity estimation generally involve measurement of arrival times of broadband pulses. A marker on the pulse waveform, such as a zero crossing, is often used. Variations in sound-speed estimates may arise from frequency-dependent attenuation and dispersion which alter spectral characteristics of waveforms and shift locations of markers. Theory is presented to correct for this distortion for Gaussian pulses propagating through linearly- attenuating, weakly-dispersive media. The theory is validated on 21 human calcaneus samples in vitro using diagnostic frequencies for bone sonometry. While the effects of disherison can be shown to be small, variations in velocity estimates due to frequency-dependent attenuation have substantial magnitude relative to the difference in average sound speeds between normal and osteoporotic bone.

We are developing quantitative descriptors of breast lesions in order to provide reliable, operator-independent means of non-invasive breast cancer identification. These quantitative descriptors include lesion internal features assessed using spectrum analysis of ultrasonic radio-frequency (RF) echo signals and morphometric features related to lesion shape. Internal features include quantitative measures of 'echogenicity,' 'heterogeneity,' and 'shadowing;' these were computed by generating spectral-parameter images of the lesion and surrounding tissue. Spectral-parameter values were generated at each pixel in the parameter image using a sliding-window Fourier analysis. Lesions were traced on B-mode images and traces were used in conjunction with spectral parameter values to compute echogenicity, heterogeneity, and shadowing. Initial results show that no single parameter may be sufficiently precise in identifying cancerous breast lesions; the results also show that the use of multiple features can substantially improve discrimination. This paper describes the background, research objective, and methodology. Clinical examples are included to illustrate the practical application of our methodology.

This paper discusses the design, fabrication, testing, and simulated imaging performance of high frequency linear arrays. Both a 2 - 2 PZT composite array with a fine spatial scale and a PbTiO₃ array have been investigated at 30 MHz. The composite array demonstrated a seven-fold increase in sensitivity over the PbTiO₃ array, as well as increased bandwidth and reduced crosstalk. The electrical impedance magnitude of the composite array was 56 ohms at 30 MHz, and the measured insertion loss was -14 dB. Simulated results demonstrate excellent lateral and axial resolution when imaging a phantom using a synthetic aperture approach. A 35 MHz device is also under development. An interconnect method using a flex circuit and sputtered metal films is used to electrically connect to each element. A curve fitting technique was then used to characterize elements of the array. Electromechanical coupling coefficients from 0.55 to 0.62 and clamped relative permittivities ((epsilon) ₃₃^S,/(epsilon) ₀) from 1200 to 2000 were observed.

A 96-channel phased array probe using Pb(Zn_1/3Nb_2/3)O₃-PbTiO₃ (PZNT 91/9) single-crystal-polymer composites has been successfully fabricated to realize broader bandwidth property. Wafers of PZNT 91/9 single crystals larger than 15 X 20 mm were prepared by a Bridgman method. 1 - 3 composites were fabricated using an improved dice-fill technique. Fabricated 1 - 3 composites with 67% volume fraction of single crystal showed electromechanical coupling factors (k₃₃') of over 80% acoustic impedances of 10 to 12 Mrayls, and dielectric constants of 1,800 to 2,100. A phased array probe with a center frequency of 3.5 MHz was fabricated using the 1 - 3 composites. A flexible printed circuit was connected to the composite and all elements were successfully cut using an improved fabrication process. The bandwidth of the PZNT 91/9 single-crystal composite probes is twice as broad as that of conventional PZT ceramic probe. Echo amplitudes of these PZNT 91/9 single-crystal composite probes are almost the same as those of conventional PZT ceramic probes.

Two dimensinal (2D) array transducers have become of great interest in the last few years, in view of the possibility of real time volumetric ultrasonic imaging. However, due to low signal to noise ratio and to the limited number of channels on available imaging systems, both sensitivity and resolution of such array are lower than those of 1D arrays. First, new high dielectric permittivity PNNZT piezoceramics are characterized and compared to classical PZT. 2D array elements are then manufactured and their experimental performances are compared. PNNZT allows an increase in element pulse echo sensitivity around 6 dB as compared to PZT array elements. The effects of the pitch and layout on the sparse array radiation pattern for several steering angles are investigated. Pseudo-random layouts are shown to have satisfactory acoustic noise level as compared to periodic layouts (vernier). Moreover, such configurations allow the pitch to be increased slightly over the classical half wavelength phased-array value, thus increasing the active area (i.e. sensitivity). Optimized array configuration leads to an increase in acoustic sensitivity of at least 6 dB and a decrease of acoustic noise level around 10 dB.

The transducer driving function for Bessel beam has circular symmetry and can be generated by annular or 2-D arrays. In 2-D array, the elements are divided into a number of rings. Taking advantage of the circular symmetry, it was shown that arranging the elements in a hexagonal pattern instead of ordinary rectangular pattern could produce almost the same field pattern with 14% less elements. Our aim here is to eliminate some of the elements of the hexagonal array and obtain a hexagonal sparse array while maintaining the quality of the generated field. In our proposed method, starting from the outer most ring, a specific number of the elements of the ring are randomly selected and turned off. The field pattern of the resulting sparse arrays is simulated and compared to the field of the array with all of its elements active. If the relative mean square error is lower than a specific threshold value, more elements of the ring are turned off. This procedure is then repeated for the next ring until reaching the central ring. Our simulations for hexagonal sparse arrays show that for an error threshold of 4%, an acceptable Bessel beam can be generated only with 22% of the transducer elements used in the original hexagonal arrays. Generated beam still shows its non-diffracting property over a limited distance.

A novel method for multi-dimensional velocity estimation using phased arrays is introduced. The method is based on the 'skew- focus' beam recently discovered by the author. The skew-focus can be steered and focused like any normal phased array beam, except that the phase fronts within the resolution cell are slanted off the main direction of propagation. Using two complementary skew-focus beams with skew angles on both sides of the main beam direction, one can obtain two independent estimates of the velocity vector. The true velocity vector can be estimated from these measurements in a straightforward manner. Skew-focus beams can be optimally synthesized in the CW case using a technique previously introduced by the author. It was discovered that the CW synthesis leads to a family of apodization functions largely independent of frequency and are directly applicable in the wideband case. These apodization functions can be expressed analytically and evaluated numerically for a given slant angle (in addition to the steering angle and focal depth). It is important to note that the resulting apodization functions are smooth and do not result from an ill-posed synthesis procedure. Therefore, these solutions are robust to quantization and discretization effects.

We have developed a method for distinguishing benign from malignant focal liver lesions based on multiscale texture analysis. In this method, ROIs extracted from within the lesions were decomposed into subimages by wavelet packets. Multiscale texture features were calculated from these subimages based upon a single-scale feature defined on the original ROIs. An artificial neural network (ANN) was used for combining these multiscale features for classification of lesions, and its performance was measured by the area under the receiver operating characteristic curve (A_z). A subset of the multiscale features that yields the highest performance is selected in a step-wise manner as the wavelet packet decomposition is performed. Three single-scale features, i.e., entropy, root mean square, and first moment of the power spectrum, are used to generate the multiscale texture features. In an analysis of 193 ROIs consisting of 50 hemangiomas (benign lesions), 69 hepatocellular carcinomas, and 74 metastases (both malignant lesions), the multiscale features yielded a high A_z value of 0.92 in distinguishing benign from malignant lesions, whereas the single-scale features yielded only 0.70. Our multiscale texture analysis method can effectively differentiate malignant from benign lesions, and thus can increase the accuracy of diagnosis of focal liver lesions in sonography.

The in vivo determination of the composition of gallstones is of interest to physicians, as it would offer the basis for choosing the appropriate therapy: high calcium content stones would call for laparoscopy, calcium free stones are smashed with a lithotriptor and flushed out using oral solvents. In publications regarding gallstone examinations employing ultrasound, opinion is divided as to whether or not the B-mode image is correlated with the stone composition. The following work aims at resolving this controversy. B-scans were examined in vivo and in vitro. The B-scans were segmented and several texture based features were tried out. Calcification was verified by x-ray imaging. The evaluation of the texture parameters of the segmented areas do not show any calcification dependent clustering. Analysis of images obtained in vitro using a water bath do not offer any better results than those obtained in vivo. However, the features proposed in the literature do show some correlation with the surface geometry. The investigation of tissue characterization using ultrasound imaging seems to be still open.

In ultrasonic tissue characterization the small reflections originating from the scattering structures inside the tissue are analyzed. To obtain diagnostic performance for tissue characterization by means of analysis of echocardiographic images we use methods of mathematical texture analysis. We investigate whether myocardial changes effect the texture of ultrasonic images and if this could be described using quantitative texture analysis. The texture analysis was computed in a single window of an ultrasound image/sequence covering the inner myocardial septum. Parameters from gray level histogram, co-occurrence matrices, run length statistics and run difference, from power spectrum and fractal dimensions were investigated to provide satisfying and generalizable results for classification of the myocardium. A set of parameters that could discriminate between normal and pathological myocardium were extracted. The results of 142 biopsies were compared with those of texture analysis in echocardiograms of 106 patients suspected having myocarditis. Using the reduced set of parameters the best sensitivity was 89.0% and the specificity was 83.6%. Myocarditis is associated with echocardiographic texture alteration. Texture analysis with methods of digital image processing can reliably identify myocarditis. A suitable solution for a computer-assisted non- invasive support for the diagnosis and detection of myocarditis was found.

Conventional B-scan systems only use the amplitude information of the backscattered signals for imaging. By imaging the local frequency dependent relative backscatter coefficient it is possible to improve the image contrast and to reduce system effects. Based on spectral analysis of rf echo signals, a procedure was developed to correct for system specific effects and to determine the relative backscatter coefficient. A new image with improved contrast results from grayscale or color coding of the frequency components of the relative backscatter coefficient. The method was applied to in vivo measurements of human prostate and transplanted kidney. For cancerous prostate tissue the relative backscatter coefficient is about 8 dB lower than for normal tissue regions. The results of the investigations on kidneys show no correlation to the current function of the organ. Certainly the different course of the frequency dependence of the relative backscatter coefficient of renal cortex and calices regions allows a contrast improvement. The method provides a system independent imaging procedure with improved image contrast for tissues with different scattering behavior and slightly reduced spatial resolution. Imaging the relative backscatter coefficient will not substitute the conventional B-mode image, but it is a useful tool providing additional information about the tissue state.

One technique of elasticity imaging, elastography, uses cross- correlation between two ultrasound A-lines to obtain an axial strain image of a sample. Usually, great care is taken with respect to the assumption that the response of the sample is elastic (lossless). In this paper, we relax this assumption and extend elastography to estimate the time-varying displacement and strain status of small samples (of the order of 1 mm). Results are presented for gel phantoms and articular cartilage samples, and they are consistent with the current theories of poroelastic materials. For example, an effective Poisson's ratio of approximately 0.5 obtained at ramp completion indicates volume conservation since the ramp time was much shorter than the characteristic relaxation time of the material. Subsequent reduction in effective Poisson's ratio coincident with stress-relaxation confirms poroelastic mechanisms whereby fluid exudation dissipates internal fluid pressurization. Observed slower relaxation of strain at the center of the sample is also compatible with these types of models. Preliminary data obtained with articular cartilage also shows valuable potential of this technique to investigate tissue biomechanics.

Three displacement estimation algorithms that can be used in ultrasonic strain estimation are introduced. Different displacement models are used in these algorithms with the intention to improve the spatial resolution and extract as much strain distribution information as possible from the recorded pre- and post-compression rf echo waveforms. The performance of these algorithms is compared by means of measuring their modulation transfer function (MTF) and generalized noise equivalent quanta (NEQ).

We are developing a method that uses acoustic radiation force to image the stiffness of the vitreous body and other soft materials. This approach applies acoustic radiation force through a series of ultrasonic pulses to generate small displacements in tissue. Motion tracking techniques are used to measure the resultant displacement. This process can be repeated at a number of locations to acquire data for image formation. A series of acrylamide phantoms were constructed to test the proposed method. Phantom speed of sound and attenuation have been characterized and found to be close to that of the human vitreous. In this paper, we present acoustic radiation force images, which clearly distinguish phantoms of differing gel concentration. We also show time-displacement curves, which indicate a viscoelastic response for this material. The images presented show that acoustic radiation force can be used to image tissue mechanical properties including displacement, relative elasticity and relative viscosity. We present data that indicates maximum displacement is linearly proportional to the power transmitted by the system. Optical data was also collected to enable visualization of the displacement field.

The large elasticity contrast possible with strain imaging promises new diagnostic information to augment x-ray, MRI, and ultrasound for the detection of tumors in soft tissue. In the past, we described the design of an elastographic system using the Fourier crosstalk concept introduced by Barrett and Gifford. The diagonal of the crosstalk matrix is related to the pre-sampled modulation transfer function (MTF) of the strain image. Another approach to measuring the spatial resolution of an elasticity image employs a linear frequency- modulated (chirp) strain pattern imposed upon a simulated ultrasonic echo field to study the strain modulation over a range of spatial frequencies in the image. In experiments, high contrast inclusions positioned at varying separations were imaged to apply the Rayleigh criterion for resolution measurement. We measured MTF curves that fell to 0.2 at a spatial frequency of 0.5 mm^-1 to 1 mm^-1 under realistic conditions. The spatial resolution for ultrasonic strain imaging strongly depends on the transducer properties and deformation patterns applied to the object. Experiments with tissue-like phantoms mimicking the properties of early breast cancer show that 2 mm spheres three times stiffer than the background can be readily resolved. Thus, the potential for using elasticity imaging to detect early breast cancers is excellent.

Complex molecular signaling heralds the early stages of pathologies such as angiogenesis, inflammation, and cellular responses to mechanically damaged coronary arteries after balloon angioplasty. In previous studies, we have demonstrated acoustic enhancement of blood clot morphology with the use of a nongaseous, ligand-targeted acoustic nanoparticle emulsion delivered to areas of thrombosis both in vitro and in vivo. In this paper, we characterize the early expression of tissue factor which contributes to subsequent arterial restenosis. Tissue factor is a 42kd glycoprotein responsible for blood coagulation but also plays a well-described role in cancer metastasis, angiogenesis, and vascular restenosis. This study was designed to determine whether the targeted contrast agent could localize tissue factor expressed within the wall of balloon-injured arteries. Both carotid arteries of five pigs (20 kg) were injured using an 8 X 20 mm angioplasty balloon. The carotids were treated in situ with a perfluorocarbon nanoparticle emulsion covalently complexed to either specific anti-tissue factor polyclonal F(ab) fragments (treatment) or non-specific IgG F(ab) fragments (control). Intravascular ultrasound (30 MHz) images of the arteries were obtained before and after exposure to the emulsions. Tissue- factor targeted ultrasonic contrast agent acoustically enhanced the subintima and media at the site of balloon- induced injury compared with control contrast arteries (p less than 0.05). Immunohistochemical staining confirmed the presence of increased tissue factor at the sites of acoustic enhancement. Binding of the targeted agents was demonstrated in vitro by scanning electron microscope images of cultured smooth muscle cells that constitutively express tissue factor. This study demonstrates the concept of molecular imaging and localization of carotid arteries' tissue factor in vivo using a new, nanoparticulate emulsion. Enhancement of the visualization of the molecular expression of tissue factor could prove to be a prognostically important predictor of subsequent restenosis. Moreover, with the incorporation of specific drug treatments into the nanoparticulate contrast agent, ultrasonic molecular imaging may yield reliable detection and quantification of nascent pathologies and facilitate targeted drug therapy.

A nongaseous, ligand-targeted perfluorocarbon nanoparticle emulsion has been developed which can acoustically enhance the presence of molecular epitopes on tissue surfaces. We demonstrate the impact of incorporating perfluorocarbons with specific phase velocities into the emulsions on the acoustic reflectivity of plasma clots targeted using these nanoparticles. Porcine plasma clots were targeted in vitro with specific perfluorocarbon emulsions using anti-fibrin antibody solution (NIB 5F3). Five perfluorocarbons were investigated: perfluorohexane, perfluorooctyl-bromide, perfluorooctane, perfluorodichlorooctane, and perfluorodecalin. Ultrasonic backscatter (17 - 35 MHz) was measured at the front surface of the clots. Backscatter enhancement was determined by comparison with untreated clots. The magnitude of enhancement depended on the perfluorocarbon emulsion used. Perfluorohexane and perfluorooctane exhibited the greatest enhancement relative to untreated clots (23 dB) and perfluorodecalin the least (18 dB), consistent with predictions from a simple acoustic transmission-line model. We conclude that targeted, nongaseous perfluorocarbon contrast agents can significantly increase the sensitivity of ultrasonic detection of low-scattering biological media, and that further optimization of these contrast agents can be realized by judicious choice of the emulsified perfluorocarbon.

The speed of sound varies with tissue type, yet commercial ultrasound imagers assume a constant sound speed. Sound speed variation in abdominal fat and muscle layers is widely believed to be largely responsible for poor contrast and resolution in some patients. The simplest model of the abdominal wall assumes that it adds a spatially varying time delay to the ultrasound wavefront. The adequacy of this model is controversial. We describe an adaptive imaging system consisting of a GE LOGIQ 700 imager connected to a multi- processor computer. Arrival time errors for each beamforming channel, estimated by correlating each channel signal with the beamsummed signal, are used to correct the imager's beamforming time delays at the acoustic frame rate. A multi- row transducer provides two-dimensional sampling of arrival time errors. We observe significant improvement in abdominal images of healthy male volunteers: increased contrast of blood vessels, increased visibility of the renal capsule, and increased brightness of the liver.

We have previously developed instrumentation for performing thermoacoustic computed tomography (TCT) of the human breast using 434 MHz radio waves. Recently, we have modified our original TCT scanner design in a number of important ways. We have increased the number of ultrasound detectors and decreased their size, and we have replaced our single RF wave- guide with a phased array of eight wave-guides. These modifications have led to increased spatial resolution, increased imaging field of view, and decreased scan time. Here we report the design considerations that led to these improvements.

A new method of acquiring and processing angular scatter data in ultrasonic imaging is presented. This method, based on the translating apertures algorithm, eliminates system-dependent changes in the received point spread function (psf) that are associated with more conventional methods of angular scatter measurement. This ensures that changes in the received echo are dominated by changes in the angular scattering behavior of insonified targets, and allows for the development of a variety of new imaging methods. Comparison of received echoes acquired at multiple interrogation angles serves to enhance the contrast of targets exhibiting variations in angular scattering behavior relative to the surrounding medium. Emphasis is placed on the improved ability to highlight biological targets that exhibit significant variations in compressibility or density relative to background tissue (e.g. breast microcalcifications, calcified atherosclerotic plaques). Simulation indicates the enhancement of breast microcalcification contrast by 10 - 30 dB over standard b-mode acquisition at 10 MHz. More sophisticated imaging methods involving the frequency dependence of angular scatter and angular speckle coherence are also discussed. Practical implementation and evaluation of this method on a modern imaging system is discussed, and expectations for the performance and utility of this algorithm in clinical diagnosis are investigated.

Since the invention of the stethoscope, the detection of vibrations and sounds from the body has been a touchstone of diagnosis. However, the method is limited to vibrations whose associated sounds transmit to the skin, with no means to determine the anatomic and physiological source of the vibrations save the cunning of the examiner. Using ultrasound quadrature phase demodulation methods similar to those of ultrasonic color flow imaging, we have developed a system to detect and measure tissue vibrations with amplitude excursions as small as 30 nanometers. The system uses wavelet analysis for sensitive and specific detection, as well as measurement, of short duration vibrations amidst clutter and time-varying, colored noise. Vibration detection rates in ROC curves from simulated data predict > 99.5% detections with < 1% false alarms for signal to noise ratios >= 0.5. Vibrations from in vivo arterial stenoses and punctures have been studied. The results show that vibration durations vary from 10 - 150 ms, frequencies from 100 - 1000 Hz, and amplitudes from 30 nanometers to several microns. By marking the location of vibration sources on ultrasound images, and using color to indicate amplitude, frequency or acoustic intensity, new diagnostic information is provided to aid disorder diagnosis and management.

Cardiac boundary extraction on echocardiographic images is essential for quantification of cardiac function. Tracing the endocardial boundaries on the end-diastolic and end-systolic images allows the computation of clinically important measures such as ejection rate. It is a clinical need for automatically detecting the borders. In this paper, we proposed a new approach for cardiac boundary extraction on echocardiographic images by directed graph. In this approach, we spread the cardiac image in the circular direction. The spread image is mapping to a directed graph. The shortest path is found by the dynamic programing algorithm. From the implemented results, we can obtain pretty good approximation for cardiac boundary extraction.

In nonlinear ultrasound imaging the images are formed using the second harmonic energy generated due to the nonlinear nature of finite amplitude propagation. This propagation can be modeled using the KZK wave equation. This paper presents further development of nonlinear diffractive field theory based on the KZK equation and its solution by means of the slowly changing profile method for moderate nonlinearity. The analytical expression for amplitudes and phases of sum frequency wave are obtained in addition to the second harmonic wave. Also, the analytical expression for the relative curvature of the wave fronts of fundamental and second harmonic signals are derived. The media with different nonlinear properties and absorption coefficients were investigated to characterize the diffractive field of the transducer at medical frequencies. All expressions demonstrate good agreement with experimental results. The expressions are novel and provide an easy way for prediction of amplitude and phase structure of nonlinearly distorted field of a transducer. The sum frequency signal technique could be implemented as well as second harmonic technique to improve the quality of biomedical images. The results obtained are of importance for medical diagnostic ultrasound equipment design.

Intravascular images from patients undergoing coronary angioplasty were obtained by a 20 MHz catheter probe. Texture analysis was performed computing features of different regions of interest, representing soft and calcified plaque and thrombus. For each class about 100 feature sets were disposed, computed in regions selected in 30 images. Texture features were classified using Bayesian classifier and a neural back propagation network. The statistical classifier led to a good discrimination between soft and calcified plaque whereas half of the thrombus feature sets were recognized as soft plaque. The accuracy of the classification result when using the neural network classifier was 87% for calcified plaque, 88% for soft plaque, and 76% for thrombus. The neural classification process was implemented as a visualization routine for PC supported classification. For this purpose the 51 texture parameters were calculated and sent to the recall routine which delivered the neural network classification result. The classification result were color encoded with red, blue and green labels for calcified plaque, soft plaque and thrombus.

As a prerequisite to performing minimally-invasive spinal surgery (MISS) with technology-guided therapy (TGT), researchers at Vanderbilt University have proposed to mathematically align the physical space of the patient with preoperative images through a surface-based registration. In order to support closed-back spinal surgeries, we have selected a non-invasive, portable imaging modality for obtaining intra-operative images, namely ultrasound (U/S). The preliminary work for the application of TGT to spinal cases has been performed on a spine phantom, scanned with an optically-tracked U/S transducer. The lumbar vertebral surface was extracted from the U/S images, and the surface pixels were converted into 3D physical-space coordinates. This set of U/S surface points was divided into a test set and a target set to be used in registration and error measurement, respectively. The test set of U/S points was registered to segmented CT spinal images of the same phantom spine using a modification of the Besl-McKay Iterative Closest Point algorithm. In a qualitative analysis of the registration, the results look favorable. The U/S points closely align with the corresponding CT surface in every image slice. By incorporating TGT into minimally-invasive spinal surgeries, the procedures are expected to yield reduced injury to normal spinal tissue and hence quicker recovery time for the patient.

The diffraction in the acoustic field of an ultrasound transducer can be modeled as the result of the interference of edge and plane waves generated from the periphery and the center of the piezoelectric element, respectively. Our objective in developing ultrasound transducers with apodized piezoelectric ceramic discs was to generate acoustical fields with reduced edge waves interference. Transducers were built with apodized ceramic discs (polarized more intensively in the central region than in the edges) and their mapped acoustic fields showed a distinct pattern when compared to those of conventional transducers. A polynomial equation describing the nonlinear poling field intensity, was used with the Rayleigh equation to simulate the nonuniform vibration amplitude distribution generated by the apodized transducers. Simulated acoustic fields were compared to experimental field mappings. The results of simulations and experimental tests showed reduction in the lateral spreading of acoustic fields produced by apodized transducers, compared to those produced by conventional transducers. The reduced presence of the lateral lobes in the apodized acoustic field is due to the minimized vibration of the disc periphery. The numerical and experimental results were in good agreement and showed that it was possible to reduce acoustic field diffraction through nonlinear polarization of the piezoelectric element.

We have developed an equipment using ultrasound transducers to help in the diagnosis of osteoporosis. The equipment consists of an X-Y axes displacement system controlled by a microcomputer and uses two ultrasound transducers in opposite sides to inspect the calcaneus region of the patient. We have used two pairs of transducers with 500 kHz and 1 MHz central frequencies. Each pair of transducers was fixed in the X-Y displacement system submerged in a small water tank with a support for the foot of the patient. The transmitter was excited with pulses of 400 - 600 kHz or 800 - 1200 kHz and the ultrasound waves propagating through the bone in the calcaneus region are received by the opposite transducer, amplified and acquired in a digital oscilloscope. The data are transferred to the microcomputer and the ultrasound attenuation and the ultrasound transmission velocity are determined. The system was tested in patients, selected from a group that had already been diagnosed using a DEXA equipment. The results showed that there is a decrease in the ultrasound transmission velocity and the ultrasound attenuation in osteoporotic patients when compared to healthy patients of the same sex and age group. The conclusion is that ultrasound attenuation and the transmission velocity in the calcaneus region may be used as parameters in the evaluation of osteoporosis using our new system.

In this paper we propose a low-cost computational method applied to the smoothing of surfaces reconstructed from noisy data, as it is typically the case in ultrasound imaging. It makes use of tracking ideas. Our method is, to our knowledge, a novel and competitive alternative to those which make use of traditional methods of optical flow for the smoothing of the normals of an object's surfaces. Those methods, as it is well- known, are very involved in calculations. Our method is based on a Kalman filter; we propose a stochastic dynamic model which exploits the spatial coherence present in the data. We end up having more efficient computational scheme with performance close to the optical flow method. A provision is made to impede the filter to diverge when the data depart from the assumed model. Our results both with synthetic and real volume data show that our proposal is realistic in terms of rendering: a good trade off between computational resources and graphical results seem to be achieved.

A beamformer is described that is used to process the data from a high frequency linear array with frequencies up to 75MHz. This unit is one part of a test system, which has the capability to characterize ultrasound transducer arrays up to 128 elements. This beamformer consists of three stages: delay unit, delay switches and summing unit. They have the following characteristics: the delay and summing units are realized in the analog domain, utilizing 4-zone dynamic focusing, and dynamic aperture switching with very wide bandwidth. The delay unit consists of a set of fixed delays with 5 ns resolution, and a set of adjustable delay lines with a resolution of 1 ns. A set of 4:1 multiplexers is used to switch the different delays for different focusing zones. In our system, the focal range is divided into 4 zones consisting of a near, middle, far and fourth zone with a different set of delays for each zone. Dynamic aperture switching is accomplished by sequential selection of the multiplexer network. The aperture is set according to the focusing zone. For near zone, middle zone, far zone and fourth zone focusing, apertures are set to 8,12,16 and 16 elements respectively. In the last stage, very wide band summing amplifiers are used with surface mount resistor networks to decrease phase errors. Test signals have been generated. The multiplexing system and delay lines have been characterized. The beamformer lateral and axial resolution have been characterized by a simulated phantom utilizing the FIELD program. The future work is to interface the beamforming architecture with a high frequency array.