Session: 2A NEW ULTRASOUND SYSTEMS Chair: H. Ermert University of Bochum 2A-1 10:30 a.m. TISSUE HARMONIC IMAGING WITH IMPROVED TEMPORAL RESOLUTION D. J. NAPOLITANO*, C. H. CHOU, G. W. MCLAUGHLIN, T. L. JI, L. Y. L. MO, and B. D. DEBUSSCHERE, ZONARE Medical Systems. Corresponding e-mail: djn@zonare.com Generally in conventional tissue harmonic imaging, a transmit beam of sufficient focusing is employed to generate levels of tissue second harmonic that provide acceptable SNR for imaging. To achieve higher temporal resolution, a small number, i.e., 5-7, of broad transmit beams are used to insonify an acoustic region and, in combination with many receive beams reconstructed from each broad transmit beam, an image is constructed. However, the level of tissue harmonic generated may be insufficient to produce an image with acceptable SNR. By employing large time-bandwidth product coded excitations, afforded by the reduced peak pressures and the lower overall number of transmit firings associated with this broad transmit beam imaging approach, sufficient SNR can be achieved. This, in combination with pulse inversion applied to receive channel data to reduce fundamental scattering contributions, has been demonstrated to produce sufficient SNR and bandwidth for tissue harmonic imaging at much higher acquisition rates than conventional systems, important when imaging dynamic structures. This approach potentially has other useful applications such as contrast agent imaging where reduced peak pressures and transmit field uniformity are important for agent integrity. Received tissue harmonic signal levels between a focused, and a broad, weakly focused transmit beam will be compared for a 3.5 MHz curved array. Spatial and temporal tradeoffs between a similar channel count, commercially available, conventional imaging system, and a broad transmit beam imaging system, with the aforementioned techniques, are discussed. Both phantom and anatomical images produced from a broad transmit beam imaging system will be shown. These images demonstrate that broadband tissue harmonic imaging with good SNR can be achieved by using a broad, weakly focused transmit beam with multiple receive beams at very high acquisition rates, thus reducing potential temporal artifacts while not compromising line density. 2A-2 10:45 a.m. BROAD BEAM COLOR FLOW IMAGING L.Y.L. MO*, T.-L. JI, C.-H. CHOU, D. J. NAPOLITANO, G. W. MCLAUGHLIN, and B. D. DEBUSSCHERE, Zonare Medical Systems, Inc. Corresponding e-mail: lmo@zonare.com 6
A broad transmit beam technique for real-time color flow imaging is described. Broad beams are generated by a 1D transducer array. Each broad beam insonifies a zone that is equivalent to 20-50 focused transmit beams, such that the full field of view can be scanned using only 3-5 firings, times the flow sample count (FSC) required for color flow estimation. For each firing, a coded waveform (up to 40 carrier cycles) is used to compensate for the lack of transmit focusing gain. On receive, the channel domain RF data is demodulated, digitized, decodedandaccumulatedinmemory,andthentransferredtoadsp-basedimage formation system, which performs dynamic receive focusing, clutter filtering, mean velocity/energy estimation and scan conversion. The much reduced data acquisition time per image frame enables longer and more rapid firing sequences (e.g. FSC=20) to be used in conjunction with different codes, including FM chirp, Barker and Golay codes. It is shown that FM chirp provides the most flexible signal bandwidth control, and Gaussian-like envelope shaping can be used for range lobe suppression. For low flow detection, a transmit sequence is proposed which consists of firing coded waveforms at higher, non-uniform pulsed repetition frequencies (PRF) of up to 4 times the user-selected PRF. Provided that the shortest PRI (=1/PRF) is larger than the 2-way transit time, additional SNR gain can be realized by receive low-pass filtering (across firings). To minimize DSP computational load, successive n-tuplets (n=2,3,4#) of the over-sampling transmit sequence are received, decoded, averaged (low-pass filter) and decimated prior to image formation. Alternatively, the entire extended FSC is processed by the DSP system, which allows more optimal low-pass filters to be combined with the high-pass clutter filter. Simulated axial responses for different signal schemes are presented which demonstrate the expected SNR gains and range lobe levels. Various sets of 3.5 MHz curvilinear and 7.5 MHz linear array color flow channel domain data have been acquired and processed both off-line and in real-time. Selected color velocity images (and image loops) of phantom and of human blood flow are presented. 2A-3 11:00 a.m. PROGRAMMABLE ULTRASOUND SYSTEM AND ITS APPLICATIONS IN RESEARCH R. MANAGULI* and Y. KIM, University of Washington. Corresponding e-mail: ykim@u.washington.edu We have designed an ultrasound back-end using off-the-shelf programmable mediaprocessors. This ultrasound system, unlike several other portable ultrasound machines, does not make any compromises in image quality, supported modes or frame rates, and meets the real-time requirements of many clinical applications. Inputs to the mediaprocessors are 16-bit RF I and Q data and outputs are the processed images. With this system, we have obtained more than 40 frames/s for B-mode (with input B data of 384 x 512 and an output image of 640 x 480) and more than 20 frames/s for color-flow mode (with 384 x 512 and 128 x 512 x 8 for B and color-flow data, respectively, and an output image of 640 x 480). Because it is programmable and has access to RF I and 7
Q data and processed data at any stage of the signal processing paths, it can support new algorithms and/or applications based on RF or mode-specific-data without any modification in hardware or signal flow path. We have been conducting algorithm/application research using many types of data sets available on this system. We have incorporated B-mode data based 3D ultrasound and achieved more than 40 volumes/s rendering speed for a 128 x 128 x 128 volume. After developing and validating a new adaptive wall-filter algorithm for improved blood-flow velocity estimation using I and Q color-flow data, it was incorporated into this system readily. We are also in the process of supporting strain imaging in real time using RF I and Q data. Our results demonstrate that the mediaprocessor-based ultrasound system provides image quality and modes that are comparable to high-end machines and at the same time offer a flexible platform for quickly integrating new applications and algorithms into the ultrasound machine, which would reduce the time required to bring innovative ideas from the research laboratory into clinical use. In this paper, we will describe this system in detail. We will elaborate the system specifications, corresponding computing requirements, programmable processors used, and system architecture. We will also describe several applications and their porting into the ultrasound system in greater details. 2A-4 11:15 a.m. A NEW COMBINED OPEN RESEARCH PLATFORM FOR ULTRASOUND RADIO-FREQUENCY SIGNAL PROCESSING A. ACQUAFRESCA 1,E.BIAGI 1, R. FACCHINI 1,HFONFARA 2,M.H.HOSS 2, R. M. LEMOR 2, L. MASOTTI* 1, S. MAZZANTI 1, A. RICCI 1, M. SCABIA 1,P. K. WEBER 2,andH.J.WELSCH 2, 1 University of Florence, Italy, 2 Fraunhofer- Institut Biomedizinische Technik, St. Ingbert, Germany. Corresponding e-mail: masotti@ingfi1.ing.unifi.it Technical researchers involved in the field of ultrasonic diagnosis need to be in full control of all settings and parameters in their equipment. In this work a research system for the acquisition and real-time processing of Radio-Frequency (RF) ultrasonic signals is presented. The system constitutes a complete solution for dealing with ultrasonic RF signals, starting from probe and beam former, down to acquisition, processing, storage and visualization. The potential of this system is realized through the consistent and synergic combination of two platforms developed by two research groups: DIPHAS (Digital Phased Array System) realized by Fraunhofer IBMT, St. Ingbert, Germany and FEMMINA (Fast Echographic Multi-parameter Multi-Image Novel Apparatus) realized by the Ultrasound and Non Destructive Laboratory, University of Florence, Italy. The main strengths of this combined system are: 1) Direct availability of raw RF data. 2) Real-time operation mode for both acquisition, processing and visualization of RF data. 3) Capability to completely control beam former parameters and excitation waveform. 4) Open and modular architecture that allows 8
the user to develop any proprietary algorithm directly in C++ and to embed them in the processing engine. The DIPHAS system is a digital beam former with 64 (expandable up to 128) transmit-receive-channels, which can support any phased array as well as linear array probe. The settings of the beam former (focusing, steering, waveform, multibeam) can be programmed freely via the PC. All acquired RF-data is transmitted to the FEMMINA system using an optical link interface. FEMMINA is a completely digital platform constituted by high speed real time processing section that integrates the architecture of a Personal Computer. These two interconnected platforms are managed by a totally programmable open and user friendly dedicated software environment. The multi-parameter extraction and the multi-image capabilities allow a continuous monitoring and control of all the system functionalities during the experimental activity. 2A-5 11:30 a.m. A PORTABLE, LOW-COST, HIGHLY INTEGRATED, 3D MEDICAL ULTRASOUND SYSTEM M. I. FULLER*, T. N. BLALOCK, J. A. HOSSACK, and W. F. WALKER, University of Virginia, Charlottesville, VA. Corresponding e-mail: mikefuller@virginia.edu Recent advances in commercial silicon integrated circuit (IC) technology have yet to be fully exploited in the development of novel 3D medical ultrasound architectures. We present a design, with prototype component experimental results, for a portable, low-cost 3D ultrasound system that utilizes innovative, state-of-the-art IC topologies and beamforming algorithms to condense the beamforming process into a small custom IC coupled with a generic, commercially available DSP chip. Furthermore, our system includes a low-cost, fully sampled 2D array placed directly adjacent to the receive circuitry allowing precise impedance matching and improved SNR. An approach consisting of a spatially broad transmit beam in conjunction with parallel receive beamformers was chosen to adequately interrogate a 3D volume. The size, cost, speed, and complexity constraints of the receive circuitry necessary to implement such an approach required a new system topology. An original transmit/receive switching scheme was developed to allow the inclusion of simple, on-chip switching elements to shunt the transmit pulse to the common node of the transducer array. This approach provided effective isolation of the high voltage transmit pulse, thus allowing the rest of the circuit to use low-voltage, low-cost commercial CMOS processing. These switching elements were experimentally shown to successfully shunt a 16 ma peak current resulting from a 100 V transmit pulse with a 25 ns risetime and a transducer element capacitance of 4 pf. The preamplifier design, consisting of two active-load differential stages, provides adjustable low-frequency roll-off, variable gain up to 85 db and an equivalent input noise of only 5 nv/ Hz in the band of interest. Following the preamplifier is the sample-and-hold stage which captures in-phase and quadrature components of the signal using a novel I/Q sampling technique. This technique reduces the digitization rate to the PRF, 9
allowing for the implementation of a compact, low-power, parallel distributed A/D converter. The A/D in each channel requires only a comparator and a latch, while a DAC and counter are implemented once and shared among the channels. This work was supported in part by the Carilion Biomedical Institute. 2A-6 11:45 a.m. THE TIME REVERSAL KALEIDOSCOPE: A NEW CONCEPT OF SMART TRANSDUCERS FOR 3D IMAGING G. MONTALDO*, D. PALACIO, M. TANTER, and M. FINK, Laboratoire Ondes et Acoustique, E.S.P.C.I., C.N.R.S UMR 7587, Université Paris VII. Corresponding e-mail: gabriel.montaldo@loa.espci.fr The design of two dimensional arrays for real time 3D imaging is a major challenge in medical ultrasound. Several thousand of transducers are typically needed to achieve the beam focusing and steering in a large 3D region of interest. Here we report a completely new approach for producing 3D images with a very small number of transducers using the combined concepts of time reversal mirrors and chaotic reverberating cavities. A small number of transducers glued on the surface of a solid cavity are used in a time reversal mode. If one face of the cavity is in contact with a fluid medium (like the body), ultrasonic waves transmitted by the transducers may take benefit of cavity reverberations to focus in any point of the fluid by using time reversal techniques. Due to the reverberations inside the cavity, waves emitted by each transducer are reflected hundred of times, creating at each reflection virtual transducers that can be observed by any observer located in the fluid. The result of such an operation is that a small number of transducers (typically 32) is multiplied by a number greater than hundred to create a kaleidoscopic transducer array presenting equivalent performances than conventional 2D matrices made of thousands of transducers. The time reversal technique allows us to use these virtual transducers to focus with a very weak side lobe level on each point of the fluid. First experimental 3D images obtained with a pioneer prototype will be presented. Beyond the scope of 3D medical imaging, this combination of time reversal processing with chaotic reverberating cavities leads to the concept of a smart transducer. 10