Acr in Doppler Hedrik

You might also like

Download as pdf
Download as pdf
You are on page 1of 7
16 OMS 11:16-22 JanuaryFebruary 1995 ee Autocorrelation Detection in Color Doppler Imaging A Review W. R. HEDRICK, PxD D. L. HYKES, PxD From the Radiology Departments of Aultman Hospital, Canton (Ohio, and St. Joseph Hospital and Health Center, Lorain, Ohio, and the Northeastern Ohio Universities College of Medicine, Rootstown, Ohio. Reprint requests: W. R. Hedrick, PhD, Radiology Department, Ault ‘man Hospital, 2600 Sixth Street, S.W., Canton, OH 44710. Autocorrelation is the most common method used to estimate velocity in color flow imaging. Measurement of reflector velocity is obtained by manipulating the echo-induced signals gener- ated from consecutive transmitted pulses. The processing steps in autocorrelation, including quadrature detection, gating, fixed echo cancel- lation, and mean frequency calculation, are dis- cussed. Aliasing, which arises from limited sampling rate of pulsed wave instruments, is described. Key words: autocorrelation, mean frequency, color Doppler imaging, quadrature detection, fixed echo canceller, aliasing. Color Doppler or color flow imaging is a relatively new scanning mode that combines gray scale imaging with two-dimensional mapping of velocity information in real time. Frame rates are slower than those achieved by rapid B-mode scanning (considerably more computational analysis is required), but rates of 4 to 32 frames per second are possible, Motion is depicted throughout the field of view by superimposing different colors on the two-dimensional, gray-scale image. A sin- gle representative velocity at each sampling site is color coded by hue or intensity of color. Real-time imaging imposes a time constraint on data collection and signal processing. Real time implies that the image must be updated every 0.05 to 0.1 seconds (corresponding to a frame rate of 10 to 20 images per second). Velocity information must be obtained for a large number of sample volumes throughout the field of view in a very limited amount of time. Autocorrelation detection satisfies this requirement. Autocorrelation, by rapidly processing the Doppler signals, was the first commercial method to achieve real-time imaging of blood flow. Because the Doppler signal is characterized with a single parameter, usu- ally the mean frequency, informational content is sacrificed, but processing time is reduced. For each depth segment along a line of sight, the mean fre- quency is converted to velocity via the Doppler shift equation. Most color flow scanners in use today em- ploy the autocorrelation detection technique. To assess motion, multiple echoes from the same /AUTOCORRELATION DETECTION IN COLOR DOPPLER IMAGING / Hedtck and Hykes 17 — 7 gage t In phase. Tete aagenence Pow otptae vane ures ouapearute FAG. Block diagram of the quadrature detection (QD) circuit reflector must be collected using a series of transmit- ted pulses. As an analogy, a series of stop-action photographs of a moving car allows one to determine the velocity. A single photograph in the series does not indicate whether or not the car is moving. Color flow imaging requires positional information as well as velocity of the moving reflector. Spatial origin of the echoes is obtained by gating the detected signals. Quadrature Detection Velocity determination is initiated with a quadra- ture detection (QD) circuit that processes the echo- induced signals from the receiver (Fig. 1). The re- ceived signal is split into two components and each component is mixed with the reference signal (one channel 90 degrees out of phase with the other chan- nel). After filtering, the output from each channel is a complex function of forward and reverse motions as well as contributions from stationary reflectors. ‘The ‘ovo output signals are labelled in-phase and out-of- phase (or quadrature). If moving reflectors along the line of sight were continuously interrogated, the QD signals associated with those reflectors would be continuous wave- forms, as illustrated in Figure 2. However, to obtain positional information, the transmitted beam must be pulsed, which causes discrete sampling of these sig- nals (Fig. 3), ‘A succession of echoes from the same reflector are collected using sequential transmitted pulses, pro- cessed through the QD circuit, and placed in a hold circuit. The hold circuit forms the composite signals illustrated in Figure 3. In the time interval following the transmitted pulse, the output from each QD channel is segmented into different depths using se- quentially clocked gates (Fig, 4). One data point is contributed to the composite in-phase signal and to the composite quadrature signal at each depth. Mul tiple transmitted pulses allow the build-up of the time-dependent QD signals (Fig. 5). The time be- tween data points is equal to the pulse repetition period (PRP). This scheme allows QD signals to be generated from many depth segments during the dwell time for the line of sight. The depth sampling interval can be 0.5 mm or smaller. In QD, the time-varying output from each channel is a complex function of amplitude and phase of the echo signals. Both stationary and moving reflectors contribute to this waveform. The phase of the re~ ceived signal from a stationary reflector is constant, whereas the phase of the received signal from a mov- ing reflector fluctuates with time. Consequently, sampling with another pulsed sound wave at a later time introduces a change in the signal level at the point in the QD output corresponding to the depth of a moving reflector (Fig. 5) A plot of QD channel output from successive trans- mitted pulses, segmented according to depth, indicates that moving reflectors produce signals with varying ‘magnitude, as shown in Figure 6. In this example, four point segments of the composite in phase signal at three different depths are shown. Two time measurements are required to obtain the information: the time follow ing the transmitted pulse assigns depth; successive transmitted pulses at time intervals of the PRP contrib- ute additional data points to the composite QD signal at each depth. Autocorrelation Autocorrelation is an intercomparison between ‘measurements acquired from the same reflector using pulsed wave ultrasound. Both channels of the QD circuit provide input to the autocorrelation detector. Processing of the echoes received from multiple depths is performed concurrently by manipulating the stream of echoes along the entire scan line ob- tained from two consecutive pulses. The delay circuit holds the QD channel outputs by a time interval equal to the PRP. This delay places the echo-induced signals from successive transmitted pulses on the same time scale, so that depth is designated by the lapsed time following the transmitted pulse. The in 18 JOURNAL OF DIAGNOSTIC MEDICAL SONOGRAPHY January/February 1995 VOL. 11, NO. 1 E IN PHASE "Sonat SIGNAL . . % . . ° . . 7 ¢ . o. cy 7 oe QUADRATURE QUADRATURE SIGNAL SIGNAL . . . - - oe oe . . . . . . . . oe oe TIME ———>> FIG. 2 _ In phase signal and quadrature signals from the quadra- ture detection circuit induced by moving reflectors. These signals are not identical but are mathematically related. phase and quadrature signals are divided into seg- ments by depth. For each depth segment, the i phase and quadrature channel outputs from consec- tutive samplings are combined by multiplication and addition and then summed in the integrator with the results from previous samplings. Registers store the computational results while the data are accumulated for one line of sight (Fig. 7).*° TIME — FIG. 3 Pulsed wave sampling depicts the in phase and quadra- ture signals in Figure 2 as a series of points For each depth two registers hold the output from the autocorrelation detector. At the conclusion of sampling along the line of sight, mean frequency and phase are computed at each depth from the values stored in registers corresponding to that depth. The phase is equal to the arctangent of the ratio of the register values, and velocity is equal to the phase divided by the sampling time interval, the PRP. Vari- ance can also be determined. [AUTOCORRELATION DETECTION IN COLOR DOPPLER IMAGING / Hedrick and Hykes SIGNAL DURING PRP JU \ lentes caret z= z= & cave ° = A FIG. 4 Gating of the quadrature detection channel output fol- lowing each transmitted pulse allows the signal to be segmented according to depth. Each quadrature detection channel output is sated. A minimum of three observations is required to determine the mean frequency. Generally, each scan line is sampled 4 to 10 times, but can be sampled as many as 32 times. The term packet size or ensemble length describes the number of pulses used to inter- rogate a single color line of sight. Large packet size (long integration time) provides the highest color def- inition (most accurate frequency estimates), but the Jong dwell time lowers the frame rate. ‘The dwell time required for each color line of sight, using autocorrelation is now examined. The frame rate (FR) in frames per second (fps) is determined by 1 FR BRP Na a where N is the number of lines per frame (Ipf), and nis the packet size. For a scanning range of 8 cm (PRP of 104 is) under the conditions of 15 fps and 64 Ipf, a packet size of 10 pulses is used to sample each line of sight. The dwell time is 1.04 milliseconds. At pulse repetition frequencies of 5 kHz or higher auto- DEPTH 12345 PULSE 1 PULSE 2 PULSE 3 PULSE 1 FIG. § Quadrature detection channel output from four succes- ‘sive pulses along the same line of sight are presented. The line of sight is partitioned into five depth segments. Motion is detected by changes in the signal level at a particular time (depth). Re- flectors at depth segments 1 and 5 are stationary; reflector at depth segment 3 4s moving. 1” 20 JOURNAL OF DIAGNOSTIC MEDICAL SONOGRAPHY anvaryFebruary 1995 VOL. 11, NO. 1 SIGNAL LEVEL TIME Quadrature detection composite signal generation from collected in Figure 5. Data points for a depth segment ‘were obtained from successive transmitted pulses. = reflectors in depth segment 1; i= reflectors in depth segment 3; @ reflectors in depth segment 5 correlation provides rapid acquisition and processing of the echo data to allow real-time imaging of flow Fixed Echo Canceller Echoes from red blood cells are lower in signal level by a factor of 100 to 10,000 than echoes gener- ated by specular reflectors (tissue boundaries). These strong signals inhibit accurate velocity estimations by the autocorrelation detector. To lower the dynamic range of the QD input to the autocorrelation detec- tor, signals corresponding to stationary reflectors are eliminated. Echoes from stationary reflectors remain unchanged in successive samplings and are removed by subtracting identical echoes in consecutive sam- plings by the introduction of a fixed echo canceller. ‘The QD channel outputs along the entire scan line obtained from two consecutive pulses are manip ulated in the fixed echo canceller. The first QD channel output is delayed by a time interval equal to the PRP, inverted, and then combined with the QD channel output from the immediately succeeding pulse. The delay places the signals measured from each transmitted pulse on the same time scale so that depth is designated by the elapsed time following the transmitted pulse. The output from the fixed echo canceller contains signals attributed to moving reflec tors; signals from stationary reflectors have been climinated (Fig. 8), Discussion ‘The widespread acceptance of color Doppler im- ng indicates that this scanning modality offers unique applicability in the clinical environment. By displaying the two-dimensional spatial distribution of the velocities and the temporal changes in these ve locity patterns, evaluation of regional flow through cout the field of view is now possible. The presence of flow, direction of flow, characteristics of flow, and focal differences in velocity within the vessel can be assessed Range-gated, pulsed-wave Doppler spectral analy- sis requires a relatively long sampling time for each line of sight, and consequently cannot satisly the design criterion of high frame rate necessary for color Doppler imaging, The detection method of autocor- relation overcomes the difficulty by rapid collection and processing of flow information. An average Dopp- ler shift is displayed at each color pixel, however. ‘The maximum Doppler shift could be presented, but this parameter is more sensitive to noise and spectral broadening. The distribution of velocity components must be obtained by another method, such as fast Fourier transform spectral analysis, Autocorrelation requires a relatively long pulse du- ration with a narrow bandwidth. Variance in the mean frequency estimate is increased when the pulse is shortened to improve axial resolution. Attenuation and diffraction of the ultrasonic beam affect the ve- locity estimation; thus, the measured mean fre- quency of reflectors moving at the same velocity var- ies with depth. Autocorrelation is based on a pulsed sampling technique, and as such, is also subject to aliasing. If flow exceeds the velocity range set by the sonographer (the Doppler shift is greater than one-half the pulse repetition frequency), then color aliasing occurs. A high-frequency shift above the Nyquist limit is inter- preted incorrectly as a lower frequency shift in the opposite direction. Aliasing causes the motion to be improperly represented by a color level corresponding to reverse flow. At very high velocity (more than twice the Nyquist limit), flow is depicted with a slow velocity color level, but the direction of movement is correct. /AUTOCORRELATION DETECTION IN COLOR DOPPLER IMAGING / Hedrick and Hykes 21 Bast ° =} RS FIG. 7 Block diagram of the autocorrelation detector. The in phase and quadrature channel outputs from two consecutive transmitted pulses are segmented by depth and then manipulated in a series of multiplication and addition steps along two pathways. An integrator in each pathway sums the computational results at each depth from repeated samplings and stores them in registers. ourrut FROM SECOND PULSE SIGNAL FROM MOVING REFLECTOR cumee a oe } revere} — ‘oureur FROM FIRST PULSE FIG. 8 Manipulation of successive quadrature detection channel output signals along the same line of sight by the fixed echo canceller. ‘The first channel output signal is delayed, inverted, and then combined with the subsequent channel output signal atthe adder. Signals from stationary reflectors are eliminated JOURNAL OF DIAGNOSTIC MEDICAL SONOGRAPHY anwaryFebruary 1995 VOL. 11, NO. 1 References 1 Foley WD, Erickson SJ. Color Doppler flow imaging. ATR Am J Roentgenol 1991;156:3-13 Hedrick WR, Hykes DL. Color Doppler imaging: imaging parameters and image quality. JDMS 1992:8:180~186. Merritt CRB. Doppler color flow imaging, J Clin Ultrasound 1987515:591-597 Mitchell DG. Color Doppler imaging: principles, limitations, artifacts. Radiology 1990;177:1-10. O'Leary DH, Polak JF. Interrogating the carotids with color Doppler imaging. Diagnostic Imaging 1988;10-204-210, Hedrick WR, Hykes DL, Starchman DE. Ultrasound Physics ‘and Instrumentation. Third edition. Saint Louis, MO: Mosby Year Book, 1995, Smith H, Zagzebski J. Basie Doppler Physics. Madison, WI: Medical Physics Publishing, 1991 8. Kasai C, Namekawa K, Koyano A, Omoto R. Real time two- dimensional blood flow imaging using an autocorrelation tech- nique. IEEE Trans Sonies Ultrasonics 1985;32:458—464, 9. Barber WD, Eberhard JW, Karr SG. A new time domain technique for velocity measurements using Doppler ultra: sound. IEEE Trans Biomed Eng 1985;32:213-229, 10. Shariati MA, Dripps JH, MeDicken WN. A comparison of colour flow imaging algorithms. Phys Med Biol 1993:38: 1589-1600,

You might also like