- Email: [email protected]
A digital image analysis method for diagnostic ultrasound calibration
Ultrasonics 39 (2002) 695–702 www.elsevier.com/locate/ultras
A digital image analysis method for diagnostic ultrasound calibration R. Zdero a
a,b,* ,
P.V. Fenton c, J.T. Bryant
a,b
Human Mobility Research Centre, Apps Medical Research Centre, Kingston General Hospital, Kingston, Ont., Canada K7L-2V7 b Mechanical Engineering Department, Queen’s University, Kingston, Ont., Canada K7L-3N6 c Radiology Department, Kingston General Hospital, Kingston, Ont., Canada K7L-2V7 Received 27 November 2001; accepted 16 June 2002
Abstract Acoustic test objects are commonly used for quality assurance testing of diagnostic ultrasound machines. However, the accompanying calibration protocols rely heavily on the judgment of the sonographer, are dependent on machine settings and are semiquantitative. In the current study, two unique test objects and protocols were designed to quantitatively determine diagnostic ultrasound parameters, namely axial resolution and geometric uniformity, and lateral resolution and geometric uniformity of the ultrasound field. The effect of focal zone, signal gain, and distance from the ultrasound probe on these parameters was assessed. The investigation was performed using a typical low-frequency diagnostic unit equipped with a 7.5 MHz linear pulse–echo probe. Results underline the need to ensure that sensitivity of routine testing regimes is adequate for the measurements to be made. This study is a preliminary part of a larger project developing an ultrasound technique to be used as an engineering design tool in a nonclinical industrial setting for quality assurance testing of total knee replacements immersed in water. Ó 2002 Elsevier Science B.V. All rights reserved. Keywords: Ultrasound; Calibration; Axial resolution; Lateral resolution
1. Introduction In order to ensure that a diagnostic ultrasound unit is functioning properly, it must be properly calibrated, two of the most important features being resolution (axial and lateral) and geometric uniformity of the ultrasound field [1–3]. Lateral and axial resolution are the minimum lateral and axial distances between two points, respectively, that ultrasound can differentiate between (Figs. 1 and 2), being important because of the implications they have on image quality and the detectability and quantitative measurement of small features [1,3–8]. It is known, that typical axial and lateral resolutions for clinical ultrasound in the 1980s––like the unit presently used––were, respectively, 0.5–3 and 0.5–10 mm in room temperature water [1,8]. Modern equipment has higher resolution. * Corresponding author. Address: Human Mobility Research Centre, Apps Medical Research Centre, Kingston General Hospital, Kingston, Ont., Canada K7L-2V7. E-mail address: [email protected] (R. Zdero).
More specifically, lateral resolution is determined by the number of lines of sight in the image, scattering properties of the medium used and display resolution and is the single most important factor in diagnostic image quality (Fig. 1). Due to the nature of beam profile, lateral resolution is expected to be at least several times larger than axial resolution. Axial resolution, normally less than 3 ultrasound wavelengths [2], is frequency and pulse length dependent and may be affected by damaged transducer crystals, changes in the nature of the excitation pulse and by distance from the transducer face (Fig. 2). Geometric or dimensional artifacts are distortions of area, volume, and distance due to improper calibration for acoustic velocity changes in materials of different densities. The degree to which these artifacts are present is an indicator of spatial and dimensional uniformity in the ultrasound field and, hence, the reliability of any quantitative measurements made. Calibration techniques for ultrasound units from the 1980s were developed to measure one or more acoustic parameters. The various protocols and test targets have
0041-624X/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 1 - 6 2 4 X ( 0 2 ) 0 0 3 8 3 - 9
696
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Fig. 2. Axial resolution. Higher frequency signals are able to discern between inline objects that are closer together than lower frequency pulses. Axial resolution for a typical diagnostic unit is 0.5–3 mm, usually below three ultrasound wavelengths, although maximum resolution detectable is one wavelength.
Fig. 1. Beam profile. When an object smaller than the beam slice thickness, BT, is anywhere in the beam path (positions 2, 3, 4), the object will be displayed as having the width BT. The acoustic intensity of the resulting image will vary depending on how close the object is to the center line (CL ) of the beam. As expected, acoustic intensity and BT are zero when the same object moves completely outside the path of the beam (positions 1 and 5). Lateral resolution, as with beam profile, changes with distance from the probe.
included subjective assessments by ultrasound technologists [9,10], traditional liver scans [13], acrylic wedge objects [5,11–14], individual point reflectors [5,15–18], ladder-shaped targets [3], the rotating-arm [3,5], and electronic phantoms [2,5]. However, by far the most common approach has been the imaging of various test objects from which either qualitative assessments or quantitative measurements are made. The most popular targets are the American Institute for Ultrasound in Medicine (AIUM) 100 mm Test Object [1,2,5,11] and the tissue mimicking (TM) phantom [2,5,6,8]. These targets are basically composed of housed arrays of stainless steel wires immersed in water (AIUM), or nylon fibers, bubbles, cyst-simulating reflectors and small graphite particles immersed in a gelatin matrix (TM). Dimensional measurements are then usually made manually from the ultrasound images using the cursor capabilities of the
machine or, more recently, with the aid of commercial computer software [19]. Note, in the current study an older unit from the 1980s is used and, as such, the AIUM and similar calibration objects with water as the working medium are more suitably used for ultrasound technology from the 1970s and 1980s. For more modern equipment, which has improved resolution, it is much more common to employ TM phantoms. It is often recommended that modern clinical ultrasound units be calibrated using TM phantoms because ultrasonic beams can become defocused and the echoes much higher when using water as the working medium. However, this increase in echo intensity actually works to the advantage of this technique in its potential application to total knee replacement testing. Specifically, it becomes easier to distinguish metal and plastic surfaces when they are immersed in water, than if submerged in a more signal-attenuating biological fluid like, say, bovine serum. The advantage of using acoustic targets with this measurement approach is evident, namely portability, quick and easy on-site quality assurance testing of ultrasound machines in routine clinical settings, and relatively low cost. However, the main disadvantage is that wires and other reflectors are arranged in fixed positions and configurations, limiting the variety of measurements that can be made. In addition, the measurements obtained using these objects are semi-quantitative, in that they are at times dependent on ultrasound settings and the subjective assessment of the sonographer. However, even newly available commercial software packages that
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
report quantitative measurements really only ensure consistency or inter-test precision due to the use of computer algorithms, but not necessarily accuracy. The purpose of the current study was, first, to introduce a new imaging and analysis protocol for testing acoustic targets similar to those described above. Specifically, instead of semi-quantitative measurement of features on an acoustic target, as is often performed during clinical quality assurance testing, the present protocol quantitatively determines the centers of a target’s particles or wires and the location of surfaces. Second, the study assesses the effect of ultrasound settings (focal zone, signal gain, and distance from ultrasound probe) on the measurement of specific quality assurance parameters (lateral resolution, axial resolution, and geometric uniformity of the ultrasound field).
2. Experimental methods 2.1. General setup An ultrasound probe (7.5 MHz, linear, pulse–echo; Sonoline SL2 model, Siemens, Erlangen, Germany) was mounted onto a 2D traversing mechanism equipped with a Vernier caliper and gage for manual probe positioning. This ultrasound unit is from 1987, its resolution not being as high as that of more modern equipment. The ultrasound signal pulsated in the axial direction through the water and the object imaged (Fig. 3). The assembly was mounted onto an aluminum frame and fixed atop a tank filled with water, which sat undisturbed for at least 24 h before tests to reach room
Fig. 3. Experimental setup for all ultrasound testing. Test objects are placed in the water tank below the ultrasound probe, which is then traversed in appropriate increments over the target.
697
temperature and reduce air bubble formation. A thermometer was used to monitor water temperature (TAVG ¼ 22:1 °C). Two acoustic test objects were imaged in this study. The nylon wire object was imaged to ascertain lateral resolution and lateral spatial uniformity of the ultrasound field, whereas the wedge object was used to measure axial resolution and axial spatial uniformity. These acoustic targets were in turn positioned underwater using a 1D sliding mechanism at known distances below the probe, which was traversed manually in known increments. Ultrasound settings were adjusted accordingly and images frozen on the ultrasound display. Specific ultrasound settings were standardized (magnification scale, scan density, pre- and post-processing) whereas focal zone, transmit power and gain levels were assessed regarding their effect on the lateral and axial parameters being measured. Tests were replicated as indicated below to reduce stochastic noise. Image ‘slices’ were captured as 24-bit colour images onto a Pentium 100 computer using Digital Video Producer software (Intel Canada, Ottawa, ON, Canada), converted to 8-bit gray maps using ImageTool 1.25 (University of Texas Health Science Center at San Antonio, Texas, USA) and analyzed with SigmaScan and SigmaScanPro software (Jandel Scientific, San Rafael, CA). Using the Jandel products, a 5 5 square pixel trace was drawn manually along a line or interface of interest, yielding a pixel gray level distribution (GLD) which represented the change in reflected acoustic energy from that interface. 2.2. Lateral measurements 2.2.1. Test object Based on some test target designs from the literature survey a nylon wire test object was assembled having a series of parallel nylon wires mounted onto an acrylic block in rake-like fashion (Fig. 4). By using the known wire-to-wire spacing for the 16 wire rake, in water lateral pixel-to-mm conversion was obtained. Also, using an eight or nine wire variant of this object for estimating lateral ultrasound resolution (i.e. beam thickness), the effect of probe-to-object depth, ultrasound focal zone settings (near field and near-and-far field) and signal gain settings (minimum, medium, maximum, subjective) were evaluated. The calibration specimen was modular with the 1D traversing mechanism described above and mounted onto it. The target was traversed to several water depths (8, 10, 20, 30, 40 mm), ultrasound settings adjusted, images video captured, and analyzed as before. Image slices were only taken at the mid-point of the object. The physical dimensions of the object were measured using a tool-maker’s microscope (Optical Measuring Tools, Ltd., serial #48, England, 1946; 0.002500 resolution).
698
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Fig. 5. Lateral geometric measurements using the nylon wire target (16 wires). The ultrasound image of acoustic reflection from the wires or wires is shown (top) along with the corresponding GLD along a horizontal trace. The distance between the wires is the distance between the GLD peaks.
Fig. 4. Nylon wire target used for ultrasound lateral resolution, or beam thickness, measurements in water (a) schematic, (b) photo.
2.2.2. Ultrasound image analysis For lateral pixel-to-mm conversion in water, a typical image of the 16 wire target with its GLD is shown in Fig. 5. The distinct GLD peaks indicate nylon wire centers, the distance between them being the center-tocenter wire distance. Using the known 3 mm wireto-wire distance (avg: ¼ 3:00 mm, s:d: ¼ 0:08 mm), a lateral pixel-to-mm conversion factor was calculated. For lateral resolution, Fig. 6 shows an ultrasound image for an eight or nine wire nylon object with the corresponding GLDs. From these traces, lateral resolution was identified as the pixel distance between two valleys on either side of a peak. Taking the valleys as the boundaries of the beam, rather than using maximum slopes or a gray-depth threshold, yields a conservative estimate. The physical distance between these wires was on average 6 mm, as alternate wires were removed from the earlier 16 wire test configuration. This was important so that the echoes of each wire would not interfere with its neighbours. 2.3. Axial measurements 2.3.1. Test object An acrylic wedge test object was constructed having an internal 1.6° hollow wedge that was filled with water during tests (Fig. 7). The design was similar to others
Fig. 6. Lateral resolution measurements using the nylon wire object (eight wires). The ultrasound image of acoustic reflection from the wires or wires is shown (top) along with the corresponding GLD along a horizontal trace. The beam thickness, or lateral resolution, is the width of the energy peaks where no echo is detected.
often used for diagnostic ultrasound calibration. The test object was mounted onto a 1D traversing mechanism with a Legoâ modularity, positioned at certain ultrasound probe-to-target depths (10, 20, 30, 40 mm), imaged, video captured and analyzed. With water as the working medium, axial ultrasound resolution and axial
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Fig. 7. Wedge target used for determining ultrasound axial resolution and axial geometric uniformity: (a) schematic, (b) photo illustrating the nylon bolts holding the mating acrylic plates together.
pixel-to-mm conversion (for computer image analysis) were measured. 2.3.2. Ultrasound image analysis Fig. 8 shows a typical ultrasound image of the wedge target, with the corresponding GLDs. For obtaining axial resolution, i.e. wavelength, vertical traces (Trace 1) were taken at a series of horizontal positions (X2 ) through the wedge toward the wedge point until just before the two dark lines, representing the opposing faces of the wedge, merged. It was at this horizontal pixel location that the vertical wedge gap was equal to one ultrasound wavelength. Knowing the geometry of the wedge (X1 ¼ 3:68 mm and 1.6°), the measured distance X2 in pixels, and the lateral pixel-to-mm conversion in water (6.24 pixels/mm), the wavelength k could be calculated. For the axial pixel-to-mm conversion factor, vertical traces (Trace 2) from five different lateral locations were drawn from probe tip to the top of the object, which is at a known depth (Fig. 8). The pixel distance from peakto-peak in the GLD was then converted to millimetres. Note that Trace 1 (for axial resolution) and Trace 2 (for axial pixel-to-mm conversion factor) show similar GLDs with very distinct peaks indicating surfaces. To avoid redundancy, the GLD for Trace 2 is not shown.
699
Fig. 8. Axial resolution and geometric uniformity measurements using the wedge target: (a) ultrasound image showing the wedge cavity, (b) GLD for a vertical trace through surfaces A to D. Note that Trace 1 (for axial resolution) and Trace 2 (for axial pixel-to-mm conversion factor) show similar GLDs with very distinct peaks indicating surfaces. When surfaces B and C become distinguishable, the gap equals k, one ultrasound wavelength.
3. Results and discussion 3.1. Lateral parameters Table 1 shows results for the lateral pixel-to-mm conversion factor and lateral resolution in water (lines 1–6) as standard deviation, maximum deviation of data with respect to the mean, and the statistical p-values for the effects of axial and lateral (horizontal) positions in the ultrasound field. This was obtained by performing an ANOVA analysis using the SAS system (SAS Institute Inc., Cary, NC, USA) on the raw data. Significant difference in results was prescribed for values of p < 0:01. Lateral geometric conversion: In the case of the lateral conversion factor in water (line 1), statistical analysis shows no effect from either the axial (vertical) depth or lateral (horizontal) position of the target within the ultrasound field. This allows for the use of the single value 6.24 pixels/mm to convert from a digital image scale to real world millimetre units, indicating acoustic field uniformity. Even so, it must be recognized that the range of the data with respect to the mean can be as high as 42.7%, indicating that a large number of samples characterizes an appropriate calibration. For the
700
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Table 1 Determining ultrasound parameters Parameter tested
Units
Mean
SD
Range (%)
Main effect pþ
pþþ
Nylon wire target (1) Lateral geometric factor (subjective gain, focus F2) (2) Lateral resolution (low gain, focus F2; 6.24 pixel/mm) (3) Lateral resolution (medium gain, focus F2; 6.24 pixel/mm) (4) Lateral resolution (high gain, focus F2; 6.24 pixel/mm) (5) Lateral resolution (medium gain, focus F ; 6.24 pixel/mm) (6) Lateral resolution (subjective gain, focus F ; 6.24 pixel/mm)
pixels/mm mm mm mm mm mm
6.24 2.59 3.51 4.57 3.50 3.04
0.55 0.66 0.69 0.45 0.83 0.48
42.7 86.6 77.5 38.6 87.1 73.8
– <0.01 <0.01 – <0.01 0.01
– – <0.01 – <0.01 –
Acrylic wedge target (7) Axial geometric factor (8) Axial resolution
pixels/mm mm
6.59 0.36
0.04 0.02
2.3 11.2
– n/a
– n/a
Type I error uncorrected for all current ANOVA results shown. Corrections would move all p-values towards null hypothesis to advantage of current study. SD ¼ standard deviation. Range (%) ¼ range of data ¼ ðmaximum minimumÞ=mean 100. pþ ¼ probability for effect of vertical (axial) depth location in ultrasound field. pþþ ¼ probability for effect of horizontal (lateral) location in ultrasound field. n/a ¼ not applicable. [–] ¼ no significant difference or effect detected.
geometric conversion parameter, assessing the effect of gain and focal zone was unnecessary since changes in these variables would only alter the intensity or clarity of the GLD, but not the spatial location of the peaks and valleys. Lateral resolution: Considering the lateral resolution or beam thickness in water for changes in ultrasound gain at the same focal zone (lines 2 and 3), it is readily apparent that there is a vertical depth effect (pþ < 0:01) with the exception of high signal gain settings (line 4). This suggests that when the signal intensity is very high, there is difficulty determining beam profile as it changes with depth because of an acoustic saturation effect. This phenomenon causes the reflection intensities from the wires to become exaggerated, being similar to television glare when brightness saturates. However, horizontal location in the acoustic field produced no effect on the beam thickness at low or high gain settings. In addition, comparison of the mean values 2.59, 3.51 and 4.57 mm (lines 2, 3 and 4) shows a practically significant rise in average beam thickness as signal gain is increased. Similar trends are observed for focal zone F when changing the gain from medium (line 5) to one based on the user’s subjective assessment of good image quality (line 6). This well-known effect of gain on apparent lateral resolution is especially evident when imaging highly reflective objects (e.g. metal or plastic) immersed in a relatively anechoic medium (e.g. water) because of the large difference in reflectivity. Now consider the effect of focal zones on lateral resolution. Specifically, for a medium gain setting, changing the focal zone from far field F2 (line 3) to nearand-far field (line 5) does not show any practical effect on the lateral resolutions of 3.51 and 3.50 mm, respec-
tively, being a 0.3% different. In both cases, lateral resolution is affected by axial (vertical) depth and lateral (horizontal) location (pþ and pþþ < 0:01). In summary, for this particular ultrasound machine, the lateral resolution and beam profile change with depth, horizontal position, and markedly with increased gain settings, but is not affected practically by focal zone. In addition, the range in measurement is somewhere between 39% and 87%. This kind of variability indicates the limitations of the clinical calibration methods used for on-site quality assurance testing of diagnostic machines. However, a single 6.24 pixels-tomm conversion constant from digital image to real world scales can safely be used. 3.2. Axial parameters Table 1 shows results for the axial pixel-to-mm conversion factor and axial resolution in water (lines 7 and 8) as standard deviation, the range of the data with respect to the mean, and the statistical p-values for the effects of axial (i.e. depth) and lateral (i.e. horizontal) coordinates in the ultrasound field. This was performed using ANOVA analysis on the axial conversion factor (line 7). Significant difference in results was prescribed for values of p < 0:01. Axial geometric conversion: In the case of axial pixelto-mm conversion, it is apparent that there is no effect from horizontal location or from vertical depth. Additionally, the data range of 2.3% is of no practical concern, thereby indicating that the single factor 6.59 pixels/ mm can be used to convert digital image to real world millimetre space. This indicates geometric uniformity in the current acoustic field.
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Axial resolution: Consider now the results for axial resolution, namely wavelength measurements using the wedge target (line 8). It is clear that the wavelength detected (0.36 mm) was almost twice that expected from an ultrasound probe with a nominal frequency of 7.5 MHz (k 0:2 mm). One possible reason for this was that the axial dimension of each pixel from the digital ultrasound images of the wedge object was the inverse of the axial pixel-to-mm conversion factor, i.e. 1/6.59 pixel/ mm ¼ 0:15 mm/pixel. The manufacturer’s nominal ultrasound wavelength (k ¼ 0:2 mm) occurred somewhere between the first and second pixel, meaning that the images did not have the resolution required to detect one wavelength, k. However, this approach was adequate enough to detect an axial resolution of less than 3k, which was considered acceptable from a diagnostic perspective in the 1980s [2]. (Note: current nominal ultrasound wavelength was confirmed to be k ¼ 0:2 mm by measurements of the output frequency of the signal generator, which fires the transducer crystals.) Thus, results indicate the need to ensure minimum sensitivity requirements of the testing protocol when measuring axial resolution.
4. Conclusion (1) New calibration protocol: The current paper has presented the design and development of unique experimental test targets and imaging protocols for determining ultrasound resolutions and geometric uniformity of the ultrasound field. The use of test objects for determining axial and lateral resolutions of a diagnostic ultrasound system has in the past been performed semi-quantitatively. These routinely used protocols rely heavily on the judgment of the sonographer and are dependent on machine settings. The present study’s quantitative protocols should be explored further and used in conjunction and compared with results from new imaging software developed specifically for calibration purposes [19]. The end use of this calibration protocol is in its application for testing total knee replacements immersed in water in a non-clinical setting as an engineering design tool. This has been published elsewhere by the authors [20]. (2) Parameters tested: As expected, the ultrasound field was geometrically uniform for all vertical depths and horizontal positions tested. Lateral resolution was constant regardless of focal zone setting or horizontal location in the ultrasound field. However, it was significantly affected by both signal gain and vertical depth from the probe face, indicating that caution should be used in reporting or utilizing lateral resolution values obtained from routine quality assurance tests. Similarly, axial resolution measurements
701
using acoustic targets can be made confidently only when the measurement protocol has the required axial sensitivity.
Acknowledgements Many thanks to Gerry Saunders and David Siu (Human Mobility Research Centre, Kingston General Hospital, Kingston, ON, Canada) for their technical expertise, Dr. Mark Kester and Osteonics Corporation (Allendale, NJ, USA) for financial and infrastructure support, Hotel Dieu Hospital (Kingston, ON, Canada) for the use of ultrasound equipment, and the School of Graduate Studies (Queen’s University) for financial support.
References [1] D.L. Hykes, W.R. Hedrick, D.E. Starchman, Ultrasound Physics and Instrumentation, second ed., Mosby Year Book, St. Louis, Missouri, USA, 1992. [2] R.L. Powis, W.J. Powis, A Thinker’s Guide to Ultrasonic Imaging, Urban and Schwarzenberg, Baltimore, USA, 1984. [3] J.P. Woodcock, Ultrasonics, Adam Hilger Ltd., Bristol, England, 1979. [4] E. Alasaarela, J. Koivukangas, Evaluation of image quality of ultrasound scanners in medical diagnostics, Am. Inst. Ultrasound Med. 9 (1990) 23–34. [5] P.L. Carson, J.A. Zagzebski, Pulse echo ultrasound imaging systems: performance tests and criteria, General Medical Physics Committee Ultrasound Task Group, American Association of Physicists in Medicine (AAPM) Report no. 8, American Institute of Physics (AIP), New York, NY, USA, November 1981. [6] A. Goldstein, W. Clayman, Particle image-resolution test object, J. Ultrasound Med. 2 (May) (1983) 195–209. [7] R.A. Harris, D.H. Follett, M. Halliwell, P.N.T. Wells, Ultimate limits in ultrasonic imaging resolution, Ultrasound Med. Biol. 17 (6) (1991) 547–558. [8] D.L. Hykes, W.R. Hedrick, L.R. Milavickas, D.E. Starchman, Quality assurance for real-time ultrasound equipment, J. Diag. Med. Sonogr. 2 (1986) 121. [9] N.M. Donofrio, J.A. Hanson, J.H. Hirsch, W.E. Moore, Investigating the efficacy of current quality assurance performance tests in diagnostic ultrasound, J. Clin. Ultrasound 12 (1984) 251–260. [10] S.C. Metcalfe, J.A. Evans, A study of the relationship between routine ultrasound quality assurance parameters and subjective operator image assessment, Br. J. Radiol. 65 (1992) 570–575. [11] R.A. Banjavic, Design and maintenance of a quality assurance program for diagnostic ultrasound equipment, Semin. Ultrasound 4 (1) (1983) 10–26. [12] P. Carson, Activities of the American association of physicists in medicine and the AIUM in ultrasound instrument performance evaluation, in: Optical Instrumentation in Medicine, SPIE, vol. 127, 1977, pp. VI253–259. [13] A. Goldstein, Slice thickness measurements, J. Ultrasound Med. 7 (1988) 487. [14] A. Goldstein, B.L. Madrazzo, Slice-thickness artifacts in grayscale ultrasound, J. Clin. Ultrasound 9 (September) (1981) 365– 375. [15] K. Brendel, L.S. Filipczynski, R. Gerstner, C.R. Hill, G. Kossoff, G. Quentin, J.M. Reid, J. Saneyoshi, J.C. Somer, A.A.
702
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Tchevnenko, P.N.T. Wells, Methods of measuring the performance of ultrasonic pulse–echo diagnostic equipment, Ultrasound Med. Biol. 2 (1976) 343–350. [16] D.R. Dietz, S.I. Parks, M. Linzer, Expanding-aperture annular array, Ultrasonic Imaging 1 (1) (1979) 56–75. [17] M.M. Goodsitt, R.A. Banjavic, J.A. Zagzebski, E.L. Madsen, An automated ultrasound transducer beam profiling system, Radiology 132 (July) (1979) 220–222.
[18] G. Lypacewicz, C.R. Hill, Choice of standard target for medical pulse echo equipment evaluation, Ultrasound Med. Biol. 1 (1974) 287–289. [19] UltraIQ literature, commercial software available from Ramsoft Inc., Toronto, Ontario, Canada, 1996. [20] R. Zdero, P.V. Fenton, J. Rudan, J.T. Bryant, Fuji film and ultrasound measurement of total knee arthroplasty contact areas, J. Arthroplasty 16 (3) (2001) 367–375.
A digital image analysis method for diagnostic ultrasound calibration R. Zdero a
a,b,* ,
P.V. Fenton c, J.T. Bryant
a,b
Human Mobility Research Centre, Apps Medical Research Centre, Kingston General Hospital, Kingston, Ont., Canada K7L-2V7 b Mechanical Engineering Department, Queen’s University, Kingston, Ont., Canada K7L-3N6 c Radiology Department, Kingston General Hospital, Kingston, Ont., Canada K7L-2V7 Received 27 November 2001; accepted 16 June 2002
Abstract Acoustic test objects are commonly used for quality assurance testing of diagnostic ultrasound machines. However, the accompanying calibration protocols rely heavily on the judgment of the sonographer, are dependent on machine settings and are semiquantitative. In the current study, two unique test objects and protocols were designed to quantitatively determine diagnostic ultrasound parameters, namely axial resolution and geometric uniformity, and lateral resolution and geometric uniformity of the ultrasound field. The effect of focal zone, signal gain, and distance from the ultrasound probe on these parameters was assessed. The investigation was performed using a typical low-frequency diagnostic unit equipped with a 7.5 MHz linear pulse–echo probe. Results underline the need to ensure that sensitivity of routine testing regimes is adequate for the measurements to be made. This study is a preliminary part of a larger project developing an ultrasound technique to be used as an engineering design tool in a nonclinical industrial setting for quality assurance testing of total knee replacements immersed in water. Ó 2002 Elsevier Science B.V. All rights reserved. Keywords: Ultrasound; Calibration; Axial resolution; Lateral resolution
1. Introduction In order to ensure that a diagnostic ultrasound unit is functioning properly, it must be properly calibrated, two of the most important features being resolution (axial and lateral) and geometric uniformity of the ultrasound field [1–3]. Lateral and axial resolution are the minimum lateral and axial distances between two points, respectively, that ultrasound can differentiate between (Figs. 1 and 2), being important because of the implications they have on image quality and the detectability and quantitative measurement of small features [1,3–8]. It is known, that typical axial and lateral resolutions for clinical ultrasound in the 1980s––like the unit presently used––were, respectively, 0.5–3 and 0.5–10 mm in room temperature water [1,8]. Modern equipment has higher resolution. * Corresponding author. Address: Human Mobility Research Centre, Apps Medical Research Centre, Kingston General Hospital, Kingston, Ont., Canada K7L-2V7. E-mail address: [email protected] (R. Zdero).
More specifically, lateral resolution is determined by the number of lines of sight in the image, scattering properties of the medium used and display resolution and is the single most important factor in diagnostic image quality (Fig. 1). Due to the nature of beam profile, lateral resolution is expected to be at least several times larger than axial resolution. Axial resolution, normally less than 3 ultrasound wavelengths [2], is frequency and pulse length dependent and may be affected by damaged transducer crystals, changes in the nature of the excitation pulse and by distance from the transducer face (Fig. 2). Geometric or dimensional artifacts are distortions of area, volume, and distance due to improper calibration for acoustic velocity changes in materials of different densities. The degree to which these artifacts are present is an indicator of spatial and dimensional uniformity in the ultrasound field and, hence, the reliability of any quantitative measurements made. Calibration techniques for ultrasound units from the 1980s were developed to measure one or more acoustic parameters. The various protocols and test targets have
0041-624X/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 1 - 6 2 4 X ( 0 2 ) 0 0 3 8 3 - 9
696
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Fig. 2. Axial resolution. Higher frequency signals are able to discern between inline objects that are closer together than lower frequency pulses. Axial resolution for a typical diagnostic unit is 0.5–3 mm, usually below three ultrasound wavelengths, although maximum resolution detectable is one wavelength.
Fig. 1. Beam profile. When an object smaller than the beam slice thickness, BT, is anywhere in the beam path (positions 2, 3, 4), the object will be displayed as having the width BT. The acoustic intensity of the resulting image will vary depending on how close the object is to the center line (CL ) of the beam. As expected, acoustic intensity and BT are zero when the same object moves completely outside the path of the beam (positions 1 and 5). Lateral resolution, as with beam profile, changes with distance from the probe.
included subjective assessments by ultrasound technologists [9,10], traditional liver scans [13], acrylic wedge objects [5,11–14], individual point reflectors [5,15–18], ladder-shaped targets [3], the rotating-arm [3,5], and electronic phantoms [2,5]. However, by far the most common approach has been the imaging of various test objects from which either qualitative assessments or quantitative measurements are made. The most popular targets are the American Institute for Ultrasound in Medicine (AIUM) 100 mm Test Object [1,2,5,11] and the tissue mimicking (TM) phantom [2,5,6,8]. These targets are basically composed of housed arrays of stainless steel wires immersed in water (AIUM), or nylon fibers, bubbles, cyst-simulating reflectors and small graphite particles immersed in a gelatin matrix (TM). Dimensional measurements are then usually made manually from the ultrasound images using the cursor capabilities of the
machine or, more recently, with the aid of commercial computer software [19]. Note, in the current study an older unit from the 1980s is used and, as such, the AIUM and similar calibration objects with water as the working medium are more suitably used for ultrasound technology from the 1970s and 1980s. For more modern equipment, which has improved resolution, it is much more common to employ TM phantoms. It is often recommended that modern clinical ultrasound units be calibrated using TM phantoms because ultrasonic beams can become defocused and the echoes much higher when using water as the working medium. However, this increase in echo intensity actually works to the advantage of this technique in its potential application to total knee replacement testing. Specifically, it becomes easier to distinguish metal and plastic surfaces when they are immersed in water, than if submerged in a more signal-attenuating biological fluid like, say, bovine serum. The advantage of using acoustic targets with this measurement approach is evident, namely portability, quick and easy on-site quality assurance testing of ultrasound machines in routine clinical settings, and relatively low cost. However, the main disadvantage is that wires and other reflectors are arranged in fixed positions and configurations, limiting the variety of measurements that can be made. In addition, the measurements obtained using these objects are semi-quantitative, in that they are at times dependent on ultrasound settings and the subjective assessment of the sonographer. However, even newly available commercial software packages that
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
report quantitative measurements really only ensure consistency or inter-test precision due to the use of computer algorithms, but not necessarily accuracy. The purpose of the current study was, first, to introduce a new imaging and analysis protocol for testing acoustic targets similar to those described above. Specifically, instead of semi-quantitative measurement of features on an acoustic target, as is often performed during clinical quality assurance testing, the present protocol quantitatively determines the centers of a target’s particles or wires and the location of surfaces. Second, the study assesses the effect of ultrasound settings (focal zone, signal gain, and distance from ultrasound probe) on the measurement of specific quality assurance parameters (lateral resolution, axial resolution, and geometric uniformity of the ultrasound field).
2. Experimental methods 2.1. General setup An ultrasound probe (7.5 MHz, linear, pulse–echo; Sonoline SL2 model, Siemens, Erlangen, Germany) was mounted onto a 2D traversing mechanism equipped with a Vernier caliper and gage for manual probe positioning. This ultrasound unit is from 1987, its resolution not being as high as that of more modern equipment. The ultrasound signal pulsated in the axial direction through the water and the object imaged (Fig. 3). The assembly was mounted onto an aluminum frame and fixed atop a tank filled with water, which sat undisturbed for at least 24 h before tests to reach room
Fig. 3. Experimental setup for all ultrasound testing. Test objects are placed in the water tank below the ultrasound probe, which is then traversed in appropriate increments over the target.
697
temperature and reduce air bubble formation. A thermometer was used to monitor water temperature (TAVG ¼ 22:1 °C). Two acoustic test objects were imaged in this study. The nylon wire object was imaged to ascertain lateral resolution and lateral spatial uniformity of the ultrasound field, whereas the wedge object was used to measure axial resolution and axial spatial uniformity. These acoustic targets were in turn positioned underwater using a 1D sliding mechanism at known distances below the probe, which was traversed manually in known increments. Ultrasound settings were adjusted accordingly and images frozen on the ultrasound display. Specific ultrasound settings were standardized (magnification scale, scan density, pre- and post-processing) whereas focal zone, transmit power and gain levels were assessed regarding their effect on the lateral and axial parameters being measured. Tests were replicated as indicated below to reduce stochastic noise. Image ‘slices’ were captured as 24-bit colour images onto a Pentium 100 computer using Digital Video Producer software (Intel Canada, Ottawa, ON, Canada), converted to 8-bit gray maps using ImageTool 1.25 (University of Texas Health Science Center at San Antonio, Texas, USA) and analyzed with SigmaScan and SigmaScanPro software (Jandel Scientific, San Rafael, CA). Using the Jandel products, a 5 5 square pixel trace was drawn manually along a line or interface of interest, yielding a pixel gray level distribution (GLD) which represented the change in reflected acoustic energy from that interface. 2.2. Lateral measurements 2.2.1. Test object Based on some test target designs from the literature survey a nylon wire test object was assembled having a series of parallel nylon wires mounted onto an acrylic block in rake-like fashion (Fig. 4). By using the known wire-to-wire spacing for the 16 wire rake, in water lateral pixel-to-mm conversion was obtained. Also, using an eight or nine wire variant of this object for estimating lateral ultrasound resolution (i.e. beam thickness), the effect of probe-to-object depth, ultrasound focal zone settings (near field and near-and-far field) and signal gain settings (minimum, medium, maximum, subjective) were evaluated. The calibration specimen was modular with the 1D traversing mechanism described above and mounted onto it. The target was traversed to several water depths (8, 10, 20, 30, 40 mm), ultrasound settings adjusted, images video captured, and analyzed as before. Image slices were only taken at the mid-point of the object. The physical dimensions of the object were measured using a tool-maker’s microscope (Optical Measuring Tools, Ltd., serial #48, England, 1946; 0.002500 resolution).
698
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Fig. 5. Lateral geometric measurements using the nylon wire target (16 wires). The ultrasound image of acoustic reflection from the wires or wires is shown (top) along with the corresponding GLD along a horizontal trace. The distance between the wires is the distance between the GLD peaks.
Fig. 4. Nylon wire target used for ultrasound lateral resolution, or beam thickness, measurements in water (a) schematic, (b) photo.
2.2.2. Ultrasound image analysis For lateral pixel-to-mm conversion in water, a typical image of the 16 wire target with its GLD is shown in Fig. 5. The distinct GLD peaks indicate nylon wire centers, the distance between them being the center-tocenter wire distance. Using the known 3 mm wireto-wire distance (avg: ¼ 3:00 mm, s:d: ¼ 0:08 mm), a lateral pixel-to-mm conversion factor was calculated. For lateral resolution, Fig. 6 shows an ultrasound image for an eight or nine wire nylon object with the corresponding GLDs. From these traces, lateral resolution was identified as the pixel distance between two valleys on either side of a peak. Taking the valleys as the boundaries of the beam, rather than using maximum slopes or a gray-depth threshold, yields a conservative estimate. The physical distance between these wires was on average 6 mm, as alternate wires were removed from the earlier 16 wire test configuration. This was important so that the echoes of each wire would not interfere with its neighbours. 2.3. Axial measurements 2.3.1. Test object An acrylic wedge test object was constructed having an internal 1.6° hollow wedge that was filled with water during tests (Fig. 7). The design was similar to others
Fig. 6. Lateral resolution measurements using the nylon wire object (eight wires). The ultrasound image of acoustic reflection from the wires or wires is shown (top) along with the corresponding GLD along a horizontal trace. The beam thickness, or lateral resolution, is the width of the energy peaks where no echo is detected.
often used for diagnostic ultrasound calibration. The test object was mounted onto a 1D traversing mechanism with a Legoâ modularity, positioned at certain ultrasound probe-to-target depths (10, 20, 30, 40 mm), imaged, video captured and analyzed. With water as the working medium, axial ultrasound resolution and axial
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Fig. 7. Wedge target used for determining ultrasound axial resolution and axial geometric uniformity: (a) schematic, (b) photo illustrating the nylon bolts holding the mating acrylic plates together.
pixel-to-mm conversion (for computer image analysis) were measured. 2.3.2. Ultrasound image analysis Fig. 8 shows a typical ultrasound image of the wedge target, with the corresponding GLDs. For obtaining axial resolution, i.e. wavelength, vertical traces (Trace 1) were taken at a series of horizontal positions (X2 ) through the wedge toward the wedge point until just before the two dark lines, representing the opposing faces of the wedge, merged. It was at this horizontal pixel location that the vertical wedge gap was equal to one ultrasound wavelength. Knowing the geometry of the wedge (X1 ¼ 3:68 mm and 1.6°), the measured distance X2 in pixels, and the lateral pixel-to-mm conversion in water (6.24 pixels/mm), the wavelength k could be calculated. For the axial pixel-to-mm conversion factor, vertical traces (Trace 2) from five different lateral locations were drawn from probe tip to the top of the object, which is at a known depth (Fig. 8). The pixel distance from peakto-peak in the GLD was then converted to millimetres. Note that Trace 1 (for axial resolution) and Trace 2 (for axial pixel-to-mm conversion factor) show similar GLDs with very distinct peaks indicating surfaces. To avoid redundancy, the GLD for Trace 2 is not shown.
699
Fig. 8. Axial resolution and geometric uniformity measurements using the wedge target: (a) ultrasound image showing the wedge cavity, (b) GLD for a vertical trace through surfaces A to D. Note that Trace 1 (for axial resolution) and Trace 2 (for axial pixel-to-mm conversion factor) show similar GLDs with very distinct peaks indicating surfaces. When surfaces B and C become distinguishable, the gap equals k, one ultrasound wavelength.
3. Results and discussion 3.1. Lateral parameters Table 1 shows results for the lateral pixel-to-mm conversion factor and lateral resolution in water (lines 1–6) as standard deviation, maximum deviation of data with respect to the mean, and the statistical p-values for the effects of axial and lateral (horizontal) positions in the ultrasound field. This was obtained by performing an ANOVA analysis using the SAS system (SAS Institute Inc., Cary, NC, USA) on the raw data. Significant difference in results was prescribed for values of p < 0:01. Lateral geometric conversion: In the case of the lateral conversion factor in water (line 1), statistical analysis shows no effect from either the axial (vertical) depth or lateral (horizontal) position of the target within the ultrasound field. This allows for the use of the single value 6.24 pixels/mm to convert from a digital image scale to real world millimetre units, indicating acoustic field uniformity. Even so, it must be recognized that the range of the data with respect to the mean can be as high as 42.7%, indicating that a large number of samples characterizes an appropriate calibration. For the
700
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Table 1 Determining ultrasound parameters Parameter tested
Units
Mean
SD
Range (%)
Main effect pþ
pþþ
Nylon wire target (1) Lateral geometric factor (subjective gain, focus F2) (2) Lateral resolution (low gain, focus F2; 6.24 pixel/mm) (3) Lateral resolution (medium gain, focus F2; 6.24 pixel/mm) (4) Lateral resolution (high gain, focus F2; 6.24 pixel/mm) (5) Lateral resolution (medium gain, focus F ; 6.24 pixel/mm) (6) Lateral resolution (subjective gain, focus F ; 6.24 pixel/mm)
pixels/mm mm mm mm mm mm
6.24 2.59 3.51 4.57 3.50 3.04
0.55 0.66 0.69 0.45 0.83 0.48
42.7 86.6 77.5 38.6 87.1 73.8
– <0.01 <0.01 – <0.01 0.01
– – <0.01 – <0.01 –
Acrylic wedge target (7) Axial geometric factor (8) Axial resolution
pixels/mm mm
6.59 0.36
0.04 0.02
2.3 11.2
– n/a
– n/a
Type I error uncorrected for all current ANOVA results shown. Corrections would move all p-values towards null hypothesis to advantage of current study. SD ¼ standard deviation. Range (%) ¼ range of data ¼ ðmaximum minimumÞ=mean 100. pþ ¼ probability for effect of vertical (axial) depth location in ultrasound field. pþþ ¼ probability for effect of horizontal (lateral) location in ultrasound field. n/a ¼ not applicable. [–] ¼ no significant difference or effect detected.
geometric conversion parameter, assessing the effect of gain and focal zone was unnecessary since changes in these variables would only alter the intensity or clarity of the GLD, but not the spatial location of the peaks and valleys. Lateral resolution: Considering the lateral resolution or beam thickness in water for changes in ultrasound gain at the same focal zone (lines 2 and 3), it is readily apparent that there is a vertical depth effect (pþ < 0:01) with the exception of high signal gain settings (line 4). This suggests that when the signal intensity is very high, there is difficulty determining beam profile as it changes with depth because of an acoustic saturation effect. This phenomenon causes the reflection intensities from the wires to become exaggerated, being similar to television glare when brightness saturates. However, horizontal location in the acoustic field produced no effect on the beam thickness at low or high gain settings. In addition, comparison of the mean values 2.59, 3.51 and 4.57 mm (lines 2, 3 and 4) shows a practically significant rise in average beam thickness as signal gain is increased. Similar trends are observed for focal zone F when changing the gain from medium (line 5) to one based on the user’s subjective assessment of good image quality (line 6). This well-known effect of gain on apparent lateral resolution is especially evident when imaging highly reflective objects (e.g. metal or plastic) immersed in a relatively anechoic medium (e.g. water) because of the large difference in reflectivity. Now consider the effect of focal zones on lateral resolution. Specifically, for a medium gain setting, changing the focal zone from far field F2 (line 3) to nearand-far field (line 5) does not show any practical effect on the lateral resolutions of 3.51 and 3.50 mm, respec-
tively, being a 0.3% different. In both cases, lateral resolution is affected by axial (vertical) depth and lateral (horizontal) location (pþ and pþþ < 0:01). In summary, for this particular ultrasound machine, the lateral resolution and beam profile change with depth, horizontal position, and markedly with increased gain settings, but is not affected practically by focal zone. In addition, the range in measurement is somewhere between 39% and 87%. This kind of variability indicates the limitations of the clinical calibration methods used for on-site quality assurance testing of diagnostic machines. However, a single 6.24 pixels-tomm conversion constant from digital image to real world scales can safely be used. 3.2. Axial parameters Table 1 shows results for the axial pixel-to-mm conversion factor and axial resolution in water (lines 7 and 8) as standard deviation, the range of the data with respect to the mean, and the statistical p-values for the effects of axial (i.e. depth) and lateral (i.e. horizontal) coordinates in the ultrasound field. This was performed using ANOVA analysis on the axial conversion factor (line 7). Significant difference in results was prescribed for values of p < 0:01. Axial geometric conversion: In the case of axial pixelto-mm conversion, it is apparent that there is no effect from horizontal location or from vertical depth. Additionally, the data range of 2.3% is of no practical concern, thereby indicating that the single factor 6.59 pixels/ mm can be used to convert digital image to real world millimetre space. This indicates geometric uniformity in the current acoustic field.
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Axial resolution: Consider now the results for axial resolution, namely wavelength measurements using the wedge target (line 8). It is clear that the wavelength detected (0.36 mm) was almost twice that expected from an ultrasound probe with a nominal frequency of 7.5 MHz (k 0:2 mm). One possible reason for this was that the axial dimension of each pixel from the digital ultrasound images of the wedge object was the inverse of the axial pixel-to-mm conversion factor, i.e. 1/6.59 pixel/ mm ¼ 0:15 mm/pixel. The manufacturer’s nominal ultrasound wavelength (k ¼ 0:2 mm) occurred somewhere between the first and second pixel, meaning that the images did not have the resolution required to detect one wavelength, k. However, this approach was adequate enough to detect an axial resolution of less than 3k, which was considered acceptable from a diagnostic perspective in the 1980s [2]. (Note: current nominal ultrasound wavelength was confirmed to be k ¼ 0:2 mm by measurements of the output frequency of the signal generator, which fires the transducer crystals.) Thus, results indicate the need to ensure minimum sensitivity requirements of the testing protocol when measuring axial resolution.
4. Conclusion (1) New calibration protocol: The current paper has presented the design and development of unique experimental test targets and imaging protocols for determining ultrasound resolutions and geometric uniformity of the ultrasound field. The use of test objects for determining axial and lateral resolutions of a diagnostic ultrasound system has in the past been performed semi-quantitatively. These routinely used protocols rely heavily on the judgment of the sonographer and are dependent on machine settings. The present study’s quantitative protocols should be explored further and used in conjunction and compared with results from new imaging software developed specifically for calibration purposes [19]. The end use of this calibration protocol is in its application for testing total knee replacements immersed in water in a non-clinical setting as an engineering design tool. This has been published elsewhere by the authors [20]. (2) Parameters tested: As expected, the ultrasound field was geometrically uniform for all vertical depths and horizontal positions tested. Lateral resolution was constant regardless of focal zone setting or horizontal location in the ultrasound field. However, it was significantly affected by both signal gain and vertical depth from the probe face, indicating that caution should be used in reporting or utilizing lateral resolution values obtained from routine quality assurance tests. Similarly, axial resolution measurements
701
using acoustic targets can be made confidently only when the measurement protocol has the required axial sensitivity.
Acknowledgements Many thanks to Gerry Saunders and David Siu (Human Mobility Research Centre, Kingston General Hospital, Kingston, ON, Canada) for their technical expertise, Dr. Mark Kester and Osteonics Corporation (Allendale, NJ, USA) for financial and infrastructure support, Hotel Dieu Hospital (Kingston, ON, Canada) for the use of ultrasound equipment, and the School of Graduate Studies (Queen’s University) for financial support.
References [1] D.L. Hykes, W.R. Hedrick, D.E. Starchman, Ultrasound Physics and Instrumentation, second ed., Mosby Year Book, St. Louis, Missouri, USA, 1992. [2] R.L. Powis, W.J. Powis, A Thinker’s Guide to Ultrasonic Imaging, Urban and Schwarzenberg, Baltimore, USA, 1984. [3] J.P. Woodcock, Ultrasonics, Adam Hilger Ltd., Bristol, England, 1979. [4] E. Alasaarela, J. Koivukangas, Evaluation of image quality of ultrasound scanners in medical diagnostics, Am. Inst. Ultrasound Med. 9 (1990) 23–34. [5] P.L. Carson, J.A. Zagzebski, Pulse echo ultrasound imaging systems: performance tests and criteria, General Medical Physics Committee Ultrasound Task Group, American Association of Physicists in Medicine (AAPM) Report no. 8, American Institute of Physics (AIP), New York, NY, USA, November 1981. [6] A. Goldstein, W. Clayman, Particle image-resolution test object, J. Ultrasound Med. 2 (May) (1983) 195–209. [7] R.A. Harris, D.H. Follett, M. Halliwell, P.N.T. Wells, Ultimate limits in ultrasonic imaging resolution, Ultrasound Med. Biol. 17 (6) (1991) 547–558. [8] D.L. Hykes, W.R. Hedrick, L.R. Milavickas, D.E. Starchman, Quality assurance for real-time ultrasound equipment, J. Diag. Med. Sonogr. 2 (1986) 121. [9] N.M. Donofrio, J.A. Hanson, J.H. Hirsch, W.E. Moore, Investigating the efficacy of current quality assurance performance tests in diagnostic ultrasound, J. Clin. Ultrasound 12 (1984) 251–260. [10] S.C. Metcalfe, J.A. Evans, A study of the relationship between routine ultrasound quality assurance parameters and subjective operator image assessment, Br. J. Radiol. 65 (1992) 570–575. [11] R.A. Banjavic, Design and maintenance of a quality assurance program for diagnostic ultrasound equipment, Semin. Ultrasound 4 (1) (1983) 10–26. [12] P. Carson, Activities of the American association of physicists in medicine and the AIUM in ultrasound instrument performance evaluation, in: Optical Instrumentation in Medicine, SPIE, vol. 127, 1977, pp. VI253–259. [13] A. Goldstein, Slice thickness measurements, J. Ultrasound Med. 7 (1988) 487. [14] A. Goldstein, B.L. Madrazzo, Slice-thickness artifacts in grayscale ultrasound, J. Clin. Ultrasound 9 (September) (1981) 365– 375. [15] K. Brendel, L.S. Filipczynski, R. Gerstner, C.R. Hill, G. Kossoff, G. Quentin, J.M. Reid, J. Saneyoshi, J.C. Somer, A.A.
702
R. Zdero et al. / Ultrasonics 39 (2002) 695–702
Tchevnenko, P.N.T. Wells, Methods of measuring the performance of ultrasonic pulse–echo diagnostic equipment, Ultrasound Med. Biol. 2 (1976) 343–350. [16] D.R. Dietz, S.I. Parks, M. Linzer, Expanding-aperture annular array, Ultrasonic Imaging 1 (1) (1979) 56–75. [17] M.M. Goodsitt, R.A. Banjavic, J.A. Zagzebski, E.L. Madsen, An automated ultrasound transducer beam profiling system, Radiology 132 (July) (1979) 220–222.
[18] G. Lypacewicz, C.R. Hill, Choice of standard target for medical pulse echo equipment evaluation, Ultrasound Med. Biol. 1 (1974) 287–289. [19] UltraIQ literature, commercial software available from Ramsoft Inc., Toronto, Ontario, Canada, 1996. [20] R. Zdero, P.V. Fenton, J. Rudan, J.T. Bryant, Fuji film and ultrasound measurement of total knee arthroplasty contact areas, J. Arthroplasty 16 (3) (2001) 367–375.