Medical Ultrasound & Electronics / Signal Processing

Waveform Index Evaluation using an Electronic Injection Phantom

 D.J. Potter, X.F. Li,
Dept. of Medical Physics & Medical Engineering, Royal Infirmary of Edinburgh.

( "IEEE Engineering in Medicine and Biology" Vol 14, No.1, January/February 1995 )

© 1995 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE. This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author’s copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

  1. Abstract
  2. Medical Doppler Ultrasound
  3. Ultrasound Test Objects
  4. Electronic Injection Phantoms
  5. Principles of Operation
  6. Hardware Implementation
  7. Phantom Performance
  8. Evaluation
  9. Waveform Injection
  10. Experimental Details
  11. Evaluation of Results
  12. Conclusions
  13. Acknowledgents
  14. References
 
In Action: Injecting a Recorded waveform In Action: Continuous Wave Circuit Wiring: rear Circuit Wiring: partial front Testing the electronic injection phantom

Abstract

A versatile Electronic Signal Injection (ESI) phantom for use in Doppler Ultrasound Quality Assurance testing has been developed within the Dept. of Medical Physics and Medical Engineering, Royal Infirmary of Edinburgh, Scotland. The phantom allows injection of real, prerecorded Doppler waveforms which facilitates testing of waveform index calculation packages. An internal, discrete audio frequency generator also allows a check to be made on the accuracy of a scanners' frequency abscissa calibration.

Sideband suppression is typically in excess of 50 dB over the range 1 - 10 MHz for audio shifts of between 300 Hz and 7kHz, allowing easy identification of directional discrimination faults (channel crosstalk) for a wide variety of scanner frequencies. The feasability of performing waveform index calculation package checking as part of a quality assurance (QA) routine with this phantom is assessed.

Medical Doppler Ultrasound

Medical ultrasound can generally be catagorised into two fields, these being image generation and blood flow rate estimation using the Doppler effect. Typically a small, hand-held transducer generates a scanning directional beam of ultrasound which maps out a 2-D image. At an interface between differing tissue types, variations in densities and speed of sound causes some of the beam energy to be scattered (reflected) back towards the transducer. An image is constructed by displaying the intensity of the echo received against time for each angular increment of the beam's sweep area. Further, if the beam encounters a moving scattering source, such as blood cells, the shift in frequency of the scattered beam caused by the Doppler effect can be used to determine the velocity of the scatterer from the Doppler formula


fd =   2 v cos(q)
cs
  fo
(1.1)

where fd is the frequency shift induced in a reflected sound beam of frequency fo when it encounters a scattering source moving at velocity v at an angle q (the `Doppler' angle) to the wavefront. The speed of sound in the medium is denoted by cs.

The differing flow rate across a blood vessel combined with scattering effects between the blood cells produce a spread of frequency shifts at any one instant, so that in `Doppler' mode (frequency shift versus time) the scanner displays an entire waveform. Combining the information from the image (blood vessel diameter, beam-blood vessel interception angle) with that from the Doppler effect one can deduce the mean blood flow rate through the vessel, a figure of significant clinical importance.

Ultrasound Test Objects

The situation is, however, not as simple as is suggested above. Two sources of error can combine to produce an erroneous blood flow rate. Firstly, several assumptions have to be made about the mean tissue sound speed, the flow profile of the blood in the vessel, etc. For non-turbulant, steady flow through a straight, cylindrical vessel far from any bends, the mean blood flow rate may be deduced simply by knowing the vessel diameter and the velocity in the centre of the vessel where the blood flow is fastest (maximum frequency of the waveform). One must also be absolutely sure that the ultrasound scanner produces reliable information. The blood vessel diameter, as often measured using electronic callipers (`imaging' mode), and the frequency shift (`Doppler' mode) are of prime importance in determining the mean flow rate through a vessel and both must be accurately known.

To this end, various test objectsd or `phantoms' have been developed to test both the physical assumptions mentioned above and the accuracy of the ultrasound scanners. Flow phantoms typically circulate blood mimicking fluid (such as a glycerine and water mixture) through a plastic tube embedded in tissue mimicking material. Fine scattering particles suspended within the fluid produce a signal similar to that from blood, and the fluid is circulated by some form of pump. Although such phantoms serve well in checking the accuracy of `imaging' mode parameters (beam-blood vessel angle and vessel diameter, etc.) and flow dynamic parameters, for QA work they suffer several drawbacks. Flow phantoms are generally bulky, extremely non-portable and take a great deal of time to prepare for use ( such as the preparation of synthetic blood). Furthermore, their performance characteristics change with time and are not therefore easily reproducable. All of this combines to make them unsuitable for checking Doppler mode parameters.

A simpler idea is embodied in the `string phantom', which uses a simple rotating thread on a system of pulley wheels to act as a line scattering source1. This phantom has the advantage of portability and ease of use, and can perform frequency abscissa checking in Doppler mode i.e. rotating the string at a constant velocity corresponding to a known Doppler frequency shift (for a set string-ultrasound beam angle q) and checking the scanner waveform to ensure the displayed frequency shift matches the theoretical one. For the testing of ultrasound scanners in Doppler mode of operation, string phantoms have one main setback - the inability to produce realistic Doppler waveforms for waveform index calculation tests. Varying the string speed can produce a variable maximum-frequency envelope, although a precise specification of the envelope is often difficult without a computer controlled motor to drive the string. Even then, as will be discussed more fully later in this paper, the advantages to be gained by doing so are questionable.

Electronic Injection Phantoms

In both the flow phantom and the string phantom, one is faced with the inability to solely test the scanner electronics. For one thing, the physical dimensions of a transducer's active area introduce an additional angular dependence through the vessel-ultrasound beam angle q, a parameter of critical importance in the Doppler equation 1.1. This results in `geometrical' spectral broadening2 of an ideal single frequency into a frequency spectrum, and clearly can be a significant source of error in determining the maximum frequency of a waveform, particularly at high Doppler angles q. This effect becomes worse as the transducer is brought closer to the scattering source. Accurate determination of the Doppler angle in itself is non-trivial given that many transducers have slightly curved faces and were designed for clinical use rather than laboratory experiments. Furthermore, a good operator can only determine q with an accuracy of approximately 3o.

Thus, if one wishes to test a scanner's frequency abscissa calibration, for instance, one finds measurement errors tend to swamp any error arising from the scanner electronics alone. This problem is eliminated with a new type of phantom which is independent of angular effects, the electronic phantom.

The concept of using electronic phantoms for Doppler Ultrasound QA work was introduced by Evans in 19893. The principle of operation is to return a frequency shifted version of the ultrasonic signal received from the scanner under test using a transmit / receive transducer pair, acoustically coupled by a sculpted block of perspex. This is typically accomplished with the aid of electronic devices which perform the act of multiplication between their two inputs, and which are known as `mixers'. The whole process has been termed `Electronic Signal Injection', or ESI. The induced frequency shift of typically a few kHz is interpreted by the scanner as a Doppler shifted version of its own emitted signal. As the Doppler shift is generated electronically, no angle measurements are necessary and that principal source of uncertainty is eliminated.

Evans used a straightforward filtering technique to remove the unwanted sideband generated by the mixing process and obtain a simulated unidirectional Doppler source. If filters are to be effective they require sharp cut-off characteristics to avoid unwanted attenuation effects over the frequency range spanning the desired sideband.

More recently, Wallace4 in Newcastle has described a wideband ESI phantom capable of injecting simple waveforms such as a sawtooth frequency sweep. By using a wideband single sideband (SSB) generator they have obtained a sideband suppression ranging between 18 and 37 dB over a range of 1 - 10 MHz, for audio shifts of up to 10 kHz. Although a considerable improvement over the prototype phantom of Evans et al, the broadband nature of the Newcastle SSB generator is responsible for their rather low sideband suppression, which is disadvantageous for scanner directional descrimination assessment. There is clearly an oppertunity for improvement in the field of electronic signal injection systems.

The elimination of angular sources of error not only make electronic phantoms ideal for frequency abscissa checking, but for another highly desireable QA procedure, but one which has been neglected somewhat in the literature - the assessment of the waveform index calculation packages on Doppler scanners. As these packages expect to operate on real Doppler waveforms, the many advantages of ESI phantoms3,4 would be incomplete without the capability to inject realistic waveforms. These factors provided the motivation for the construction of an improved ESI prototype phantom during the period May 1992 to May 1993.

1.1  Principle of Operation

Figure 1.1 Figure 1.1 illustrates the principle of operation of the Edinburgh ESI phantom. The acoustic ultrasound signal from the scanner is converted into a radio frequency (RF) electrical signal by means of an input transducer attached to the ESI phantom. The signal is then split into two signals shifted in phase by [(p)/2] radians. Each RF signal is then electronically mixed with an audio frequency signal in a device known as a balanced `modulator' or `mixer', which performs the mathematical operation of multiplication. The audio frequency signals, as might be expected, are also [(p)/2] radians out of phase with each other.

The output of each modulator consists of two (radio) frequencies, termed sidebands, which are symmetrically displaced by an amount equal to the audio frequency from the scanner's frequency, termed the carrier. This technique forms the basis for AM radio transmission, from which we borrow the relevant terminology. Summation of the modulator outputs produces a single sideband (SSB) output necessary to simulate a unidirectional Doppler source.

By introducing [(p)/2] phase shift circuits at the appropriate places, the unwanted sideband is cancelled out, giving rise to the name of this technique - the `phasing method'. This technique is capable of producing very high suppression of the undesired sideband through destructive interference between two antiphase phase RF signals. One disadvantage is that this technique cannot easily be implemented over a wide bandwidth, as the phantom of Wallace et al demonstrated. Instead, we have opted to choose twelve scanner frequencies between 2 MHz and 10 MHz in use, according to a survey, in hospitals within Lothian and Fife Regions.

1.2  Hardware Implementation

Figure 1.2 In the Edinburgh phantom precise quadrature phase shifts at each scanner frequency are achieved by using parallel tuned R-L-C circuits for the ultrasonic signals, as depicted in figure 1.2. At the resonant frequency of each such circuit, two signals in quadrature phase with one another but of equal amplitude are drawn from the inductor and capacitor using high impedance buffers. The inductor is slightly variable to allow fine tuning of each circuit for precise quadrature phase shifting.

Although successful cancellation of the undesired sideband ideally requires identical respective signal amplitudes and exact 900 phase shifts, an error bound of 1o is widely regarded as an upper limit for good sideband suppression5. Sideband suppression figures suggest we have obtained a phase error between RF signals of less than 1o, in comparison with the results of Wallace et al who quote phase errors of 4o with their broadband RF filter. By using multiple parallel tuned circuits of the type shown in figure 1.2 we have successfully devised a compromise to obtain maximum sideband cancellation at frequencies of practical use to us with a minimum of electronic complexity.

Waveform injection is achieved by means of a second, audio phasing board generously donated by Duncan Cairns, manager of Dynamic Imaging Ltd, Livingston. This board uses a combination of active and passive filters to produce two signals of equal amplitude but in quadrature phase with one another from any audio input frequency in the range 300 Hz to 7 KHz. Technically, the board consists of a `polyphase' network6,7, a network of resistors and capacitors connected in a ring configuration, the output of which is a fourfold set of quadrature signals. The reader is referred to the references for further details which are out of place here.

Figure 1.3 Further flexibility is provided by an onboard audio oscillator microchip which produces quadrature phase audio signals. These may be fed directly to the mixers, allowing single frequency injection again at twelve distinct Doppler shift frequencies between 200 Hz and 14 KHz. A single switch on the phantom front panel allows selection of either the onboard audio signal generator or an external signal source (passed through the audio phasing board), such as a cassette deck, for pre-recorded waveform injection. The complete phantom is pictured in figure 1.3.

1.3  Phantom Performance

Three parameters are of interest in characterisation of the phantom's performance

  1. sideband suppression over the operational audio frequency range using a typical scanner (radio) frequency

  2. linearity of the injected audio signal power

  3. sideband suppression over the operational radio frequency range as measured using a typical audio frequency
Whilst the first and last characterisation parameters may be somewhat intuitive, the second requires a word or two of justification. Many waveform index calculation packages analyse the relative power content of a Doppler spectrum to determine a characteristic spectral index. It is therefore important that any electronic injection phantom linearly transfers the power spectral density of the input audio signal (the simulated Doppler frequency shift signal) to that of the signal returned to the scanner. sources of non-linearity may lie in amplification or mixer stages, for instance, and are beyond the scope of this paper.

Tests were performed using a Phillips PM5133 audio function generator to produce the required audio frequencies and a Wavetek function generator emulated scanner signals with a radio frequency sine wave. As waveform injection is a priority for this phantom, all audio frequencies were piped through the Dynamic Imaging audio phasing board for this experiment. The onboard audio generator typically provides a constant 30 db sideband suppression and is primarily intended for frequency abscissa calibration checking.

Evaluation

Figure 1.4 Figure 1.4 shows the results of tests one and two above. Sideband linearity is quantified by inputting an audio signal at typical operating power levels into the phantom and measuring the injected sideband power level relative to the noise floor. All power measurements were taken with the aid of a Marconi Instruments model TF 2370 graphic spectrum analyser.

As supplied, the operational range of the audio phasing board is given as between 300 Hz and 7 kHz. It will be observed that sideband suppression is optimum, however, between 750 Hz and 6 kHz over which an excellent constant suppression of 52±2 dB is achieved. The error figure arises from the resolution of the spectral analyser display screen. Over this frequency range the lower sideband power steadily increased from 61 dB to 64 dB relative to the noise floor of the spectral analyser display, illustrating a remarkably high degree of linearity over the working range of the phantom.

Figure 1.5 In figure 1.5, the sideband suppression as measured using a typical audio frequency of 4 kHz is plotted for eight several scanner frequencies spanning the ESI phantom's operational range, characterisation test 3 above. Again a very high suppression of approximately 53 dB is obtained over the operational range of the phantom. The slightly lower suppression observed at 2 MHz is due to a fractionally higher inductance used in the parallel tuned phasing circuit causing imperfect quadrature phase shifting. This has since been rectified by changing to a lower inductance coil for this filter.

Figure 1.6 Overall the phantom exceeds all previously published sideband suppression figures. The built-in audio oscillator allows 100% reproducable audio test signals to be injected for frequency abscissa calibration checking (Figure 1.6) whilst any waveform lying within the wide audio bandwidth of the phantom may be injected via the Dynamic Imaging audio phasing board.

1.4  Waveform Injection

The second half of this paper reports on the use of the Edinburgh ESI phantom as used in waveform injection mode to study the reproducability of results from the waveform index calculation package of a Doptek CW scanner. As it is not the intention to define a generally valid protocol here, specifics are discussed with relevance to the Doptek scanner used in this experiment.

In order that possible variations in waveform indices between different scanners (and waveform index calculation packages) be detectable, the unavoidable experimental error induced by making the measurement must first be quantified. The nature of this error is identified by considering the experimental procedure. A pre-recorded waveform is first injected for a set number of cycles (time period) into the scanner under test and several cycles are chosen for the waveform index calculation package to operate on, as per clinical practice. A light pen attached to the Doptek scanner is used to select the start and finish of the chosen time period over which the waveform index calculation is to be made. The effect of small differences in the start and finish times on the resulting index is quantified by calculating the mean and standard deviation of the resulting waveform index upon repetition of this procedure several times.

Other intrinsic sources of error to the waveform index calculation must be minimised in order to make useful observations of operator induced error. The waveform generated by a computer controlled phantom is characterised by a highly reproduceable frequency envelope, and is thus suitable for the minimisation of errors associated with a variable waveform envelope between cycles. The reader should also recall that statistical variations in the power spectral density within an average cycle also play a part in determining the resulting waveform index error figure, but these are hopefully smaller than those induced by operator error. As a control, the experiment was also conducted on a simple, artificially generated waveform which has a constant spectral power density at identical phases of each cycle. (The control waveform was produced by frequency modulating a 4.65 kHz audio signal with another of approximately 1.5Hz so that the resulting frequency envelope sinusoidally oscillates between 2.5 kHz and 6.8 kHz, lying within the linear operational audio bandwidth of the electronic phantom.)

Finally, it has been experimentally observed that waveform index calculation packages are influenced by scanner gain settings. For completeness the above experimental procedure was repeated for three different scanner gain settings. These gain settings were similar to those routinely used in clinical practice with this scanner, especially the `medium' gain (table 1.1), with the highest gain used still producing a noise free spectrum.

Experiment

A physiologically realistic, highly periodic waveform (figure 1.7) was obtained using a computer controlled, in-house flow phantom8 and stored on audio cassette. The phantom transducers were coupled to the Doptek transducer through a perspex block similar to that used by Evans3 and the waveform was replayed on a Yamaha KX-530 precision cassette deck at a fixed audio power output.

For each of the three gain settings used (low, medium and high as defined in table 1.1) the mean pulsatility index (PI) was determined over a five cycle period using the Doptek's [ 15/16] percentile method. This was repeated six times to gain the mean and standard deviation figures shown in table 1.1 for the real waveform, and in table 1.2 for the control waveform depicted in figure 1.8.

Doptek Gain Spectral Description P.I. s PI Figure 1.7
Low Weak, completely unsaturated - 1.940.03
Medium Slightly saturated -1.350.05
High Highly saturated - 1.33 0.04
Table 1.1: Variation of Real Waveform P.I. with Doptek Gain.


Doptek Gain Spectral Description P.I. sPI Figure 1.8
Low Weak, completely unsaturated -0.8230.004
Medium On point of saturation -0.8250.005
High Completely Saturated -0.820 0.000
Table 1.2: Variation of Control Waveform P.I. with Doptek Gain.

Evaluation

The results presented in table 1.1 suggest that use of a periodic, realistic test waveform allows highly reproducable measurements of PI for gain setting likely to be used in clinical practice. The weakest gain setting is most unlikely to be used in practice and serves to demonstrate a dependence of computed PI with gain. The control waveform exhibits markedly less variation in computed PI with gain, with the standard (measurement) deviation at least a factor of 10 times smaller than that obtained with the realistic waveform. The difference is most probably attributable to the nature of the power spectral densities of each waveform, as discussed earlier, but in any case the measurement error associated with each type of waveform is small. The entire experiment was repeated on a second Doptek scanner unit and produced PI readings consistent with those of tables 1.1 and 1.2, to rule out effects introduced by machine specific errors.

1.5  Conclusions

A portable, all electronic Doppler phantom has been described. Use of a commercially obtained wide-band audio phase shifting board allows the injection of realistic waveforms, a facility allowing novel QA techniques such as the accuracy of waveform index calculation packages to be assessed. The design of this phantom results in exceptionally high sideband supression and a highly linear power spectral density injection characteristic at a discrete number of scanner operating frequencies which lie in the range of 2-10 MHz. The Edinburgh ESI phantom can also inject any of twelve internally generated, discrete simulated Doppler shift audio frequencies for fast testing of the Doppler frequency mapping operation of ultrasonic scanners. Such tests are free from large, angle-induced measurement errors as the Doppler shift is created electronically.

The phantom has been used to demonstrate that waveform index calculation package checking is quite feasable and can obtain highly reproduceable results as demonstrated in this paper using the PI package on a Doptek CW scanner. Although the realistic waveform was produced by a flow phantom, such phantoms have difficulty in maintaining waveform integrity (average power spectral density, envelope etc.) over time and are almost completely unportable. The ESI phantom therefore also provides an essential link enabling the same `captured' flow phantom waveform to be used again and again on many different scanners with the typical ease of use (portability, speed of setup) of an electronic device.

1.6  Acknowledgments

We would like to express our deepest thanks for the advice of Mr. Tom Anderson, Mr. Dale Cameron, Dr. Harry Brash and the many staff of the Dept. of Medical Physics and Medical Engineering at the Royal Infirmary of Edinburgh for their assistance in this project. We gratefully acknowledge the assistance of the Scottish Home and Help Dept. in funding this research project.

References

  1. Walker AR, Phillips DJ, Powers JE: Evaluating Doppler devices using a moving string test target. Journ. Clin. Ultrasound 10:25-30, 1982.

  2. Bascom PA, Cobbold RS, Roelofs, BH: Influence of spectral broadening on continuous wave Doppler ultrasound spectra: a geometric approach. Ultrasound in Med. Bio. 12:387-395, 1986.

  3. Evans JA, Price R, Luhana F: A novel device for Doppler ultrasound equipment. Phys.Med.Biol.34(11):1701-1707, 1989.

  4. Wallace JJA, Martin K, Whittingham T A: An experimental single-sideband acoustical re-injection test method for Doppler systems. Physiol.Meas.14: 479-484, 1993.

  5. Radio Communication Handbook, McGraw Hill, 1957.

  6. Hosking R: Polyphase direct conversion SSB. Electronics and Wireless world: 202-206, 1994.

  7. The Art of Electronics, second edition, Cambridge University press, New York. p295

  8. McDicken W N: A versatile test object for the calibration of ultrasonic Doppler flow instruments. Ultrasound.Med.Biol. 12, 245-249, 1986.


File translated from TEX by TTH, version 3.02.
On 26 Oct 2001, 19:34.