~va Advanced Technology Laboratories, Inc., Bellevue, WA. 9R005 I. ABSTRACT

Transcription

1 COMBINED TWO-DI~mNSIONAL TISSUE/FLO~ IMAGING 1 E. Aaron Howard, 1 Harco Brandestini, 2 Jeff Powers, 1 Saeed Taheri Mark E. Eyer,1 David J. Phillips,1 and Edward B. Weiler2 1. Center for Bioengineering, University of Washington, Seattle, ~va Advanced Technology Laboratories, Inc., Bellevue, WA. 9R005 I. ABSTRACT A functional sectorscan system has been built which provides two-dimensional (2D) images based on a coherent echo/doppler processor. The phase and magnitude of the demodulated signal (COMplex video) are analyzed by means of a fast digital processor. An image field of 256 X 256 pixels is stored in Memory. A digital scan convertor and color coding scheme formats the image for standard raster display. The tissue and flow features are distinguished by contrasting color or grey scale assignments. A microprocessorbased operating system provides flexible control of sector format, pulse repetition frequency, sample clock, encoding scheme, etc. The significant features of the echo/doppler processor are: The output of the receiver is log compressed and demodulated by a quadrature pair of double-balanced mixers to preserve the phase and magnitude information within a 1 MHz bandwidth. A flash 8 bit A/D convertor is multiplexed between the "real" and "imaginary" components at a 7 MHz rate. A linear approximation (hypotenuse function) is used to calculate the magnitude term. The phase is computed by means of a look-up table. To obtain the Doppler shift for any given point, the derivative de/dt is substituted by the change of phase (~e/~t) between consecutive transmit/receive cycles. Initial in vivo trials show that the system is capable of producing a longitudinal view of the carotid artery. Both tissue and flow features are clearly distinguishable. K. Y. Wang (ed.), Acoustical Imaging Plenum Press, New York

2 534 E. A. HOWARD ET AL. II. INTRODUCTION AND BACKGROUND One of our primary objectives is the continued development of combined pulsed Doppler flowmetry and 2D sectorscan imaging. The instrument described herein (Ecomega II) is the latest generation in this development and represents a state-of-the-art device for tissue/flow imaging. Ecomega II provides two fundamental modalities for imaging. The first is a combination of 2D tissue images with superimposed flow features (B/Q mode), and the second is a combination of tissue and flow features presented in a standard time-motion display nvq mode). Both of these modalities have been reported on in earlier publications (Brandestini, et al., 1979; Eyer, et al., 1979). The emphasis in this report is on B/O mode imaging and a new, coherent echo/doppler processor. The use of pulsed Doppler techniques for describing regions of flow (Baker, 1970) has had both peripheral vascular and cardiac applications. Flow maps of vessel lumens (Hokanson, et al., 1972; Curry, 1978) have been made without the aid of 2D tissue images for orientation. While flow features have been extracted, it has not been possible to identify vessel walls or cardiac anatomy. In contrast to pulsed Doppler techniques, 2D sectorscan images have been used to visualize vascular walls and cardiac anatomy. This technique can provide important tissue visualization, but yields no hemodynamic information. Clearly, the combination of 2D sectorscan and pulsed Doppler is advantageous. The Duplex echo/doppler sectorscan systems described by Barber (1974) and Phillips (1978) served to combine the two techniques described 'above into a single instrument. These systems provided non-storable sectorscans and single sample volume pulsed Doppler flowmetry, thus allowing the user to investigate flow features with the aid of 2D tissue image orientation. The Duplex scanner went through another evolution by the addition of the digital multigate Doppler (D~m) and the digital field store (DFS) as reported by Brandestini (1978) and Eyer (1978). This new scanner was called Ecomega I and is well reported (Eyer et al., 1979, Brandestini et al., 1979, Howard et al., 1979, and Stevenson et al., 1979). Ecomega I provided the first B/O and H/Q modalities for clinical trials. Due to limitations in the instrument and exclusive 5 MHz transducer operation, Ecomega I is used most extensively in pediatric cardiology, providing M/Q images. Ecomega I served to establish the feasibility of combining 2D tissue images and 2D flow features into a superimposed 2D echo/flow image. Many of the problems inherent to such a prototype scanner

3 COMBINED TWO DIMENSIONAL TISSUE/FLOW IMAGING 535 have been improved upon in the new Ecomega II scanner described in this report. The Most notable improvement is the new coherent echo/doppler detection scheme which, in addition to being a novel method for processing echo/doppler information, provides a very versatile and sophisticated instrument for continuing our tissue/ flow imaging research. A. System III. PRINCIPLES OF OPERATION Echomega II consists of two primary subsystems: the amplitude and Doppler detector, and a display media (Fig. 1). A microprocessor provides overall control of the scanner and facilitates the necessary operator interface. The Digital Echo/Doppler Processor (DED) is the heart of the scanner and will receive the main emphasis of this report. Huch of the technology of the DED is an evolution of the Digital Hultigate Doppler (DHD). The mm established the feasibility of a multigate Doppler instrument utilizing a time multiplexed digital processor and thus served as the basis for developing the DED. The Image Field Signal Processor (IFSP) is a further development of the Digital Field Store reported by Eyer (1978). The IFSP provides digital processing of the echo/doppler data such that images may be stored, mixed, and eventually displayed by a standard color television monitor. This report will not go into a great deal of detail regarding the IFSP since it represents primarily a simplification of the DFS rather than new technology. B. Digital Controller The entire scanner is under the control of a central timing module (digital controller). This module provides all necessary clock functions. Timing is generated from a master clock ( }1Hz) with phase coherence maintained throughout the system. Although the reader might expect this central timing to be the only reasonahle approach, it should he noted that Ecomega I was not entirely synchronous. Thus, a single digital controller for Ecomega II was a significant improvement. Relevant timing signals will be discussed as they apply to the Modules described below.

4 536 TRANSMIT ~!!!!!!~~~IVE I ROTARY SCAN HEAD _--.,. E. A. HOWARD ET AL. SAMPLE & HOLD AID I DETECTOR PHASE 2 ~2 DETECTOR A : Re +Im e ~ arctan~ " ;:;~;~;Y I LINE ~UFFER I INTERPULSE DELAY AMPLITUDE ROTOR ~ SERVO I~ ISSUE IMAGE (B&W ~ DIGITAL I SCAN CONVERTER TISSUE /FLOW SECTOR SCANNER Fig. 1. Tissue/Flow Scanner Block DiagraM. The diagram illustrates the separation of the processor into a tissue (echo "E") and a flow (Doppler "D") channel after detection and digitization. Tissue and flow information is being determined simultaneously and stored into two separate banks of memory (256 X 256 X 4 each) by means of the digital scan convertor. The data is then read from the two memories in a TV line format and the two fields appear superimposed. The color coding allows combining of four dimensions (range, angle, intensity, velocity) into a composite TV picture.

5 COMBINED TWO DIMENSIONAL TISSUE/FLOW IMAGING 537 C. RF Section The DED processor is based on a phase coherent detection scheme. Gated bursts of ultrasound are transmitted via a piezoelectric transducer. The same transducer acts as both transmitter and receiver of the backscattered signal. The entire transreceiver is designed to operate at 3 or 5 ~rnz with an overall bandwidth of 1 MHz. The gated bursts are 3 cycles in duration (at the respective carrier, 3 or 5 HHz) with a pulse repetition frequency (PRF) dependent upon the depth and zoom selected for the particular application. The received signal is amplified, logarithnically compressed, and demodulated. The detection is accomplished with a pair of double balanced nixers. Quadrature reference signals are fed into the mixers and the result is a pair of quadrature outputs. This scheme is essential to the direction sensitive frequency estimator (Brandestini, 1978) and ensures that all phase and magnitude information is preserved (Fig. 2). Echo Intensity: "aml2litude" of Im A ~Re2+ 1m2 8 = arctan Re A max {I Re I, 11m I} velocity = dx d8 " 119 dt.. dt lit Fig. 2. Rectangular to Polar Conversion to Derive Hagnitl1~e and Phase from Orthogonal Projections of Complex Video Sip,nal. The complex video signal generated by synchronous quadrature detection consists of two orthogonal components. The signal contains amplitude and phase informa~ion for any point in time (range). In our system, we convert from rectangular to polar at fixed sample periods (e.g., 256 over the entire depth range). While the amplitude can be obtained during a single transmit/receive cycle (monopulse), the Doppler detection requires samples from two successive cycles (frequency = temporal derivative of phase)

6 538 E. A. HOWARD ET AL. The output of the demodulator maintains its phase relationship to the master clock for all points along the range of the sound beam. Thus, this signal is called the range phase signal. It is the processing of this range phase signal which allows us to extract both Doppler and echo simultaneously, as the signal consists of "real" and "imaginary" components. D. Sample and Hold The DED processor is a discrete-time and discrete amplitude instrument. The discrete-time format allows a single processing unit to provide 256 sequential channels of information during each transmit/receive cycle. The discrete amplitude format allows semiconductor memory storage and digital processing of the complex video signal. In order to provide 256 channels of complex video information, the DED must sample the output of the demodulator. This means that both of the quadrature outputs must be sampled prior to the analog-to-digital conversion (A/D). A pair of sample/hold circuits performs this function. At this point, the complex video is in discrete-time format. Thus, a single processor can be time multiplexed to perform the operations necessary to provide 256 channels of echo/doppler information. The next crucial step is to perform the A/D conversion necessary to utilize digital techniques for processing and storage. E. A/D Conversion Both the "real" and "imaginary" components of the sampled complex video must be A/D converted. This is accomplished by time multiplexing a flash 8 bit A/D convertor between the two components. The convertor is operating at 7 UHz. It is worth reviewing the signal thus far processed. The output of the A/D convertor is a sequential stream of data representing 256 pixels. This sequence is actually 256 pairs of discretetime, discrete amplitude data elements. Each pair can be referred to as a "real" and "imaginary" component. Since the "real" and "imaginary" components are processed sequentially in time, yet represent data acquired from the same "point" in space, it is important that they be closely spaced in time. In fact, the "real" and "imaginary" components are never more than 140 nanoseconds apart, and this represents (nominally) a spatial ambiguity of m. This exceeds the resolution capabilities of our current transducers.

7 COMBINED TWO-DIMENSIONAL TISSUE/FLOW IMAGING 539 F. Echo Detection (Hagnitude Construction) Classic echo processing in ultrasound systems utilizes asynchronous amplitude demodulating schemes such as envelope detectors. The DED extracts the echo information directly from the sampled complex video signal. In a linear system, the best representation of magnitude on a pixel by pixel basis would be: Since this operation would require excessive dynamic range and the DED performs log compression in the RF section, it is appropriate to use an approximation (hypotenuse function) to calculate the magnitude term. We have found the following function to be satisfactory: A '"!:\ax ( I RE I, I 1M I ) G. Doppler Detection In order to derive velocity (Doppler) information from the complex video signal, the processor has to determine the phase angle for every spatial sample. This is achieved by conputing the arctangent of the "imaginary" divided by the "real" portion of the signal (Brandestini, 1978). Just as velocity can be regarded as the temporal derivative of position or range, the Doppler-shifted frequency is represented by the derivative of the phase of the received signal. In our sampled processor, we substitute the derivative by the difference between two transmit/receive cycles. As shown in an earlier report, this simplification yields results very close to the true instantaneous frequency. As the phase shifted signal scattered from the moving blood cells contains a great deal of random information, temporal or spatial averaging has to be performed to get an estimate of the actual actual blood flow velocity. So far, the system averages only in time (1 20 msec per line); it has therefore been necessary to collect the 2D flow velocity map over several heart cycles (Fig. 3). In a real-time flow imaging apparatus, a certain amount of spatial averaging would have to be incorporated to compensate for the short observation time due to rapid scanning. H. Image Field Signal Processor (IFSP) The IFSP accepts image data from the DED and properly formats for B/Q or H/Q displays. The functions performed include:

8 540 E. A. HOWARD ET AL. digital scan conversion image storage color coding video interface for RGB or RS170 TV display The scan conversion is necessary to reproduce the 60 sector currently created by our Mechanical scan head. This conversion is accomplished by direct computation of the polar to rectangular equations: X 3/4 R sin + Xo YR cos + Yo where the scaler 3/4 accounts for the 3 x 4 aspect ratio of the display and Xo ' Yo are the offsets representing the "window" of interest. Images are stored digitally in high speed static RAM fields. The IFSP currently has the capability of storing four separate images with a pixel depth of 4 bits. Plans are being made to expand the pixel depth to 6 bits. Tissue and flow are stored in separate image fields and combined in the readout circuitry. The microprocessor allows the operator to select flow only, tissue only, or composite displays. Color coding schemes are designed to allow the operator to select predetermined codes to represent tissue and flow. In addition, the operator can create color schemes of his own choosing through alphanumeric keyboard control. The prime objective is to provide contrasting representations for tissue and flow. Typically, tissue is encoded in shades of grey scale while flow is represented by bi-directional color schemes. Thus, four dimensions of information (range, angle, intensity, velocity) are displayed on the TV screen. The video interface provides all of the necessary timing signals for standard TV synchronization. The primary display mode utilizes separate red, green, and blue (RGB) drives for the TV monitor, thus providing a high quality color image. Also available is an NTSC composite video channel for video tape recording. IV. PRELUUNARY RESULTS Fig. 3 is a black and white reproduction of a color image constructed by Ecomega II. It is representative of a two-dimensional B/Q mode image containing both tissue and superimposed flow features. The original color-encoded flow is reproduced here as slightly fainter grey tones within the vessel lumen.

9 COMBINED TWO-DIMENSIONAL TISSUE/FLOW IMAGING 541 ID: YOU OR ME Ct1/S ANGULAR DEPENDANCE 4.14 /12.09 R+: 0.40/.29 HD:PVX I UofW Ultrasound Seattle Fig. 3. 2D Image of the Common Carotid Artery Showing Anatomic and Superimposed Flow Features In addition to the tissue/flow image, supplemental information is provided to facilitate ease of interpretation as well as serve as a reference upon which quantitation of important parameters can be made. Along the top of the image are cm range markers which, for this case, indicate that the horizontal field of view is 6 cm at the skin surface. A color bar for the velocity information is located to the left, while a grey scale bar for detected echo amplitude is located at the right. If color were possible, one would see a "rational red" color scheme for flow velocities away from the pulsed Doppler transducer while those velocities toward the transducer would be displayed in a "rational blue scheme" ranging from dark blue, through green, and then to yellow. The phrase "rational color scheme" is intended to denote a color sequence in which the viewer can readily distinguish relative velocity magnitudes. Below the image, alphanumerics relate patient identification and various instrument settings. The 6 cm/s is the velocity magnitude step change from the flow velocity scale. The 4.14 denotes that the pulsed Doppler PRF is 4.14 KHz while in the Duplex mode (i.e., alternating pulsed echo with

10 542 E. A. HOWARD ET AL. pulsed Doppler) while the denotes the pulsed Doppler PRF in KHz in the Doppler-only mode. This change in pulsed Doppler PRF is essential if accurate velocity information is to be obtained in regions of disease where high velocities are generally present (Phillips et al., 1979). The 0.40 just after "R+:" is the time in seconds after the last R-wave of the ECG and is the time when the two-dimensional tissue image was stored. The 0.20 indicates that the flow velocity image was generated 0.20 seconds after the R-wave. The scan head used was PVX2-5 and the composite video frame stored is number The 20 image in Fig. 3 shows a longitudinal sectional of a normal common carotid artery just proximal to the carotid bulb located to the right and out of the field of view. The skin surface is seen anterior with an intervening muscle layer seen at the left between the skin and arterial wall. The two bright targets within the confines of the vessel at the left are reverberant artifacts from sound within the scan head boot. This type of reverberent artifact is readily distinguished from "real" tissue interfaces through slight movements of the scan head while observing the real-time image. During an examination, the operator surveys local anatomy and listens to the flow character at selected points by manually placing a sample volume (i.e., in Duplex mode) at various sites within the blood vessel. When a location of interest is reached, and while holding the scan head still, a foot switch is pressed. This maneuver puts the system into a Doppler-only mode which "freezes" the last entire two-dimensional tissue image and increases the Doppler PRF to 12 KHz. The pulsed Doppler transducer is then manually swept through the object volume to detect flow velocities, if present, and to display each velocity magnitude according to its proper spatial location and direction. The flow image is constructed at 0.20 seconds past the R-wave, which is near the time at which the peak velocities occur in this portion of the carotid artery. The two-dimensional flow velocity image is constructed over approximately 2.5 cm of blood vessel. Since a multi-channel pulsed Doppler is employed, only about 40 heart beats were required to generate the flow image. The multi-channel feature makes such two-dimensional flow images practical by significantly reducing the time over that required for the single sample volume approach. The "angular dependence" of the velocity magnitudes as a function of Doppler angle with respect to the vessel axis can be appreciated in this image. The magnitudes are relatively low toward the left, where an "unfavorable" Doppler angle approaches 90 0, while, when scanning toward the carotid bulb, the angle becomes more favorable and higher magnitudes are detected. At the present time, the factor of Doppler angle is not taken into account in the

11 COMBINED TWO DIMENSIONAL TISSUE/FLOW IMAGING 543 image display, but should be in future studies to provide accurate information. When combining the anatomic image with the flow velocity image, in the IFSP, differentiation of the two types of information is relatively straightforward. The hypothesis is that the integration of tissue and velocity information into a single display format can provide more information than could be acquired by either modality employed separately. The reader is referred to Brandestini et al., 1979, for color examples obtained from this system. While initial clinical trials on presumably normal blood vessels are encouraging, clinical usefulness for diseased peripheral arteries has yet to be shown. Considerable flexibility is built into this system in order to modify operating parameters and output display characteristics, as need be. Problems related to noise in the Doppler channel and obtainment of higher quality twodimensional tissue images are primary areas of future efforts. ACKNOWLEDGEMENTS The authors wish to acknowledge the technical assistance of Jay Borseth, Gordon Kirkendall, George Hahler, John Ofstad, Bob Olson, and Vern SimMons. We wish to thank Carolyn Phillips for her work in preparing this manuscript. The technical research was supported by NIH Grant HL REFERENCES Baker, D. w. (1970), Pulsed ultrasonic Doppler blood-flow sensing. IEEE Trans. Sonics Ultrasonics 17, Barber, F. E., Baker, D. W., Nation, A. W. C., Strandness, D. E., Jr., and Reid, J. M. (1974), Ultrasonic duplex echo-doppler scanner. IEEE Trans. Biomed. Engr. 21, Brandestini, M. A., Forster, F. K. (1978), Blood flow imaging using a discrete-time frequency meter. Ultrasonics Symposium Proc. IEEE Cat.D 78 CH1344-1SU. Brandestini, M. A., Eyer, H. K., and Stevenson, J. G. (1979), M/Qmode echocardiography -- the synthesis of conventional echo with digital multigate Doppler. Echocardiology, Third Symposium on Echocardiology, Rotterdam, Netherlarids, June Curry, G. R. and White, D. N. (1978), Color coded ultrasonic differential velocity arterial scanner (Echoflow). Ultrasound in Hedicine and Biology 4,

12 E. A. HOWARD ET AL Eyer, M. K. (1978), A microprocessor based digital scan converter and color display system for ultrasonic image presentation. tmsters thesis, Electrical Engineering, University of Washington, Seattle, Washington. Eyer, M. K., Brandestini, M. A., Phillips, D. J., and Baker, D. W. (1979), Color digital echo/doppler image presentation. Article submitted to Ultrasound in Medicine and Biology. Hokanson, D. E., Mozersky, D. J., Sumner, D. S., HcLeod, F. D., Jr. and Strandness, D. E., Jr. (1972), Ultrasonic arteriography: A non-invasive method of arterial visualization. Radiology 102, Howard, E. A., Brandestini, M. A., Eyer, H. K., and Weiler, E. B. (1979), Color-coded digital echo/doppler imaging, storage, and display. Proc. 4th International Symposium on Ultrasound Imaging and Tissue Characterization, pp Phillips, D. J., Blackshear, W. H., Baker, D. W. and Strandness, D. E., Jr. (1978), Ultrasound Duplex scanning in peripheral vascular disease. Radiology/Nuclear Hedicine 8, Phillips, D. J., Powers, J. E., Eyer, M. K., Blackshear, Jr., W. H., Bodily, K. C., Strandness, Jr., D. E., Baker, D. W. (1979), Detection of peripheral vascular disease using the Duplex scanner III. Article submitted to Ultrasound in Hedicine and Biology. Stevenson, J. G., Brandestini, H. A., Weiler, E. B., Howard, E. A. and Eyer, M. K. (1979), Digital multigate Doppler with color echo and Doppler display - diagnosis of atrial and ventricular septal defects. Proc. 52 Heeting of the American Heart Association, Part II, Vol. 60, #4.