Published online before print April 26, 2006, 10.1148/radiol.2393051070
(Radiology 2006;239:869-874.)
© RSNA, 2006
Anthropomorphic Breast Phantoms for Qualification of Investigators for ACRIN Protocol 66661
Ernest L. Madsen, PhD,
Wendie A. Berg, MD, PhD,
Ellen B. Mendelson, MD and
Gary R. Frank
1 From the Department of Medical Physics, University of Wisconsin, 1300 University Ave, Room 1530, Madison, WI 53706 (E.L.M., G.R.F.); Breast Imaging Consultant, Lutherville, Md (W.A.B.); and Department of Radiology, Northwestern University School of Medicine, Chicago, Ill (E.B.M.). Received June 25, 2005; revision requested August 16; revision received September 1; final version accepted September 21. Supported by grants from the National Cancer Institute (CA89008) and the Avon Foundation.
Address correspondence to E.L.M. (e-mail: elmadsen{at}wisc.edu).
 |
ABSTRACT
|
|---|
The purpose of this study was to evaluate various ultrasonic properties of breast phantoms developed for use in qualifying investigators for participation in the American College of Radiology Imaging Network (ACRIN) protocol 6666, "Screening Breast Ultrasound in High-Risk Women." Specifically, a tool was sought to consistently measure the performance of radiology personnel in detecting and characterizing lesions similar to those expected with screening breast ultrasonography (US). The phantoms are equivalent to one another except for the randomization of positions of 14 of the 17 simulated lesions. The lesions differ in depth and ultrasonic properties. Representative values of propagation speed, attenuation, relative echogenicity, and mass density are reported for all tissue-mimicking components. Beam refraction occurs at the interface between the subcutaneous fat layer and the glandular parenchyma and can result in beam distortion artifacts similar to those encountered in clinical breast US.
© RSNA, 2006
|
INTRODUCTION
|
|---|
Anthropomorphic phantoms for use in testing the performance of ultrasonographic (US) scanners and for training sonographers have been previously reported (1–7). By using production methods and materials validated in developing previous phantoms, a new set of six equivalent anthropomorphic breast phantoms was produced for use in qualifying investigators for participation in a multicenter trial of screening breast US, American College of Radiology Imaging Network (ACRIN) protocol 6666 (http://www.acrin.org). Results of the qualification task are reported in a companion article in this issue (8). Properties and depths of the 17 lesions in each phantom were designed to simulate lesions seen in clinical breast US. Thus, the purpose of our study was to evaluate various ultrasonic properties of breast phantoms developed for use in qualifying investigators for participation in the ACRIN protocol for screening breast US in high-risk women.
|
MATERIALS AND METHODS
|
|---|
Phantom Composition and Production
The materials used in the phantoms consist of water-based gelatin (200 bloom gelatin derived from calfskin; Vyse Gelatin, Schiller Park, Ill) with microscopic solid or liquid (oil) particles to provide attenuation and echogenicity (9,10). Nonfat components, such as the simulated glandular parenchyma, and most lesions, contain solid particles of graphite (catalog no. 9039; Superior Graphite, Chicago, Ill) and glass beads (45–53 µm in diameter; Potters Industries, Parsippany, NJ) dispersed in the gelatin. The fatty portions of the phantom contain oil droplets admixed with the gelatin; the oil also produces a lowered propagation speed and density. The oil is a solution of 50% safflower oil (Hollywood; Hain Celestial Group, Melville, NY) and 50% kerosene (catalog no. 329460; Aldrich Chemical, Milwaukee, Wis). A small (1% by volume) concentration of formalin (catalog no. F-79; Fisher Scientific, Pittsburgh, Pa) raises the melting point of the materials (by means of formaldehyde cross-linking) to 100°C, and a 5% concentration of 1-propanol (catalog no. A-414; Fisher Scientific) is added for preservation.
The methods used in production have been described previously (5) except that during production of the glandular region, lesions are laterally positioned precisely (within 1 mm) by suspending them on 0.3-mm-diameter stainless steel wires that are parallel to the acrylic sides of the phantom (Fig 1). The molten glandular parenchyma is then introduced around the wires and lesions and allowed to congeal, after which the wires are removed. Exact positioning is sought, as one of the tasks of the user of a phantom is to specify the position of each lesion detected.
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a: Diagrams of anthropomorphic breast phantom. (a) View from the top through the scanning window and (b) view from the side. The particular ordering of the lesions here corresponds to phantom 1. Markers around the scanning window are at 1-cm intervals for use in specifying the positions of lesions.
|
|
View larger version (34K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b: Diagrams of anthropomorphic breast phantom. (a) View from the top through the scanning window and (b) view from the side. The particular ordering of the lesions here corresponds to phantom 1. Markers around the scanning window are at 1-cm intervals for use in specifying the positions of lesions.
|
|
All but one of the lesions are spherical (ie, round). The exception is a "double-ended cone" consisting of two cones joined at their 3.5-mm-diameter bases; one cone is 4 mm in length and the other is 6 mm in length, resulting in an irregular shape. The lesions are distinct in size, echogenicity, ultrasonic attenuation, and depth (distance from the scanning window).
Diagrams of two views of one of the phantoms are shown in Figure 1. The phantom has the shape of a rectangular parallelepiped, with a depth of 6 cm, bounded by acrylic walls except for a 10.0 x 15.5-cm scanning window. The phantom simulates the breast compressed against the chest wall, the usual situation in supine US breast imaging. There is a subareolar region and a subcutaneous fat layer bonded to a distal "glandular" layer.
Lesion Location and Phantom Imaging
The location of the 17 lesions indicated in Figure 1 corresponds to phantom 1. The positions of 14 lesions (lesions 1–7 and 10–16) were randomized for each of the other five phantoms. Randomization of lesion location was performed to reduce the tendency for scanning personnel to learn the positions of specific lesions from one imaging session to the next—that is, a different phantom could be employed if a participant required a repeat analysis of lesion characteristics and depths. Three lesions were not included in the randomization because of manufacturing complications. All six phantoms were imaged (E.L.M., with 28 years of experience in phantom development) by using an Antares US unit (Siemens Ultrasound, Mountain View, Calif) equipped with a VFX 9-4 linear-array transducer operating at 9 MHz to verify that desired properties had been obtained for each lesion; representative images are shown in the Results section.
Phantom Testing
At the time of production of each component material in the phantoms, a 7.6-cm-diameter, 2.5-cm-thick test cylinder of the material was produced for measurement of ultrasonic properties (propagation speed, attenuation coefficient, and echogenicity). The cylinder of material is bounded on its curved surfaces by a 6-mm-thick acrylic wall and on the flat parallel surfaces by 25-µm-thick Saran Wrap (Dow Chemical, Midland, Mich). The methods for measuring speed and attenuation have been previously described (11). One of the authors (G.R.F., with 26 years of experience in phantom development) made speed and attenuation measurements on test samples for all six phantoms. Relative echogenicity was measured (E.L.M.) on samples corresponding to all six phantoms by comparing dual images on the Antares unit (Siemens), using the calibrated attenuator and accounting for previously measured attenuation differences. Densities were computed with knowledge of densities and volume fractions of component materials.
|
RESULTS
|
|---|
Values of relevant ultrasonic properties of the various materials in the phantoms are shown in the Table. The ultrasonic properties shown in the Table are representative of the various tissue types (12,13) except that the attenuation coefficient of simulated breast glandular tissue is much lower than reported in vitro values (12). The echogenicities of different simulated tissues are distinct. The attenuations are comparable for fat, glandular parenchyma, and those lesions intended to have no shadowing or enhancement. For "high-attenuation" lesions, the attenuation is about twice that of the glandular material, and for "very high attenuation" lesions, the attenuation is about 20 times that of the glandular material. The propagation speeds have similar values for all nonfat-mimicking materials, whereas that for fat is about 100 m/sec lower than that for the nonfat materials.
US images of representative lesions are shown in Figures 2–8. These images were made (by E.L.M.) by using phantom 2 and an Antares unit (Siemens) equipped with a VFX 9–4 linear-array transducer operating at 9 MHz. The interface between the glandular parenchyma and subcutaneous fat layer is irregular, which produces beam refraction challenges to visualization of simulated lesions in the glandular region (Fig 3).
View larger version (115K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2: US image shows double-ended cone (lesion 7) near center (open arrow), surrounded by the glandular region. The long axis of the lesion is horizontal. The scalloped interface at about 1 cm depth is that between the subcutaneous fat and the glandular region. The linear horizontal echo at the top marks the scanning window; a reverberation 2 mm below the scanning window likely arises within the transducer. The retromammary fat pad (thick solid arrow) is at about 5 cm depth, and the very echogenic pectoral muscle layer is below that. A horizontal transducer reverberation artifact (thin solid arrow) is exhibited in the retromammary fat pad.
|
|
View larger version (71K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3: US image shows 3-mm-diameter hyperechoic lesion (solid arrow) near center (lesion 4), with high attenuation relative to glandular material causing posterior shadowing. The long shadow (open arrow) distal to the cusp in the subcutaneous fat layer results from intense refractive beam distortion at the cusp.
|
|
View larger version (70K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4: US image shows 3-mm-diameter cyst (arrow) near retromammary fat pad (lesion 14). Beam refraction at the subcutaneous fat–glandular interface may be responsible for the apparent irregular shape of the cyst. Some internal echoes are seen in the cyst; these artifactual echoes probably result from presence of US beam side lobes.
|
|
View larger version (111K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7: US image shows 3- and 10-mm subtle hypoechoic lesions with attenuation equal to that of surrounding glandular region (ie, no posterior features). The 3-mm lesion (lesion 5) is on the right (solid arrow), and the 10-mm lesion (lesion 3) is on the left (open arrow).
|
|
View larger version (117K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8: US image of 6-mm cyst (solid arrow; lesion 9) and 5-mm fat sphere (open arrow; lesion 8) in subareolar region. Note specular reflections at proximal and distal ends of the fat sphere and mixed posterior enhancement and shadowing distal to the fat sphere; latter artifacts are due to beam refraction at vertical edges of the fat sphere. The fat sphere (lesion 8) was intended to be hyperechoic to the superficial fat layer and slightly hypoechoic to parenchyma but is nearly isoechoic to parenchyma in some phantoms; thus, investigators were not scored on their interpretation of echogenicity of lesion 8. Vertical linear echoes within the cyst are likely related to use of a thin (0.3-mm) stainless steel wire to position the lesion during manufacture; it was removed after congealing of tissue-mimicking glandular parenchyma surrounding the lesion and wire.
|
|
|
DISCUSSION
|
|---|
A set of anthropomorphic breast phantoms has been developed for use in US imaging with realistic simulations of normal tissue components, including fat layers, glandular parenchyma, and pectoral muscle. Representative US propagation speeds, attenuations, densities, and relative echogenicities of the various simulated tissue types are reproduced in the phantoms. Thus, refraction effects and specular reflections occur as they might in clinical breast US.
Beam refractions are created at the irregular interface of the simulated parenchyma and subcutaneous fat layers. Such refractions often occur in imaging of human breasts at the interface of fat lobules as the ultrasound propagation speed in fat is significantly lower than in nonfat tissue. Shadowing at the interface of fat lobules can be a source of false-positive findings and may require changing the angle of insonation, increasing pressure of scanning, or possibly even short-interval follow-up (ACRIN protocol 6666; available at: http://www.acrin.org). The presence of beam refractions in the phantoms facilitates demonstration of this phenomenon to users and can help increase awareness of its physical basis, although it was not possible to use increased scanning pressure as a method to eliminate such refractions in the phantom. Simulated retromammary fat pad and pectoral muscle layers lie distal to the glandular region, aiding the scanner operator in determining the relative attenuation of a lesion compared with the glandular parenchyma by amplifying enhancement and shadowing effects within the two deeper layers.
A broad range of lesion sizes and types is present for assessing the user's ability to distinguish cyst from solid tissue and to determine the size and shape of an abnormality, whether enhancement or shadowing exists, and whether a lesion is hyper- or hypoechoic.
Smaller (3-mm-diameter) cysts (ie, lesions 14–16) did not show enhancement. This is likely because the attenuation coefficient of the glandular parenchyma is too low, resulting in lower scanner gains than would be seen with clinical conditions. In future breast phantoms, the attenuation coefficient of the glandular parenchyma will be made three or four times greater to be in agreement with published in vitro values (12).
In retrospect, we regret not having included more than one nonspherical lesion in the phantoms. This oversight limits the potential for drawing conclusions regarding radiologists' ability to identify irregularly shaped lesions of different sizes. Another study limitation is that, since cancers tend to be irregular while benign lesions are more frequently round or oval and circumscribed, the phantoms may not be able to discriminate between radiologists who can detect cancers at screening US and those who cannot.
The phantoms are admittedly rather geometrically simple compared with the level of complexity found in real breasts. However, the phantoms provide a reproducible, quantitative, and realistic tool for comparing skills of users in lesion detection and recognition of important lesion characteristics similar to those expected in clinical breast US.
|
ADVANCE IN KNOWLEDGE
|
|---|
- Breast US phantoms have been constructed for systematic evaluation of operator performance.
|
FOOTNOTES
|
|---|
Abbreviations: ACRIN = American College of Radiology Imaging Network
Author contributions: Guarantor of integrity of entire study, E.L.M.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, E.L.M.; experimental studies, E.L.M., W.A.B., G.R.F., E.B.M.; and manuscript editing, all authors
See also the article by Berg et al in this issue.
Authors stated no financial relationship to disclose.
|
References
|
|---|
- Madsen EL, Zagzebski JA, Ghilardi-Netto T. An anthropomorphic torso section phantom for ultrasonic imaging. Med Phys 1980;7:43–50.[CrossRef][Medline]
- Madsen EL, Zagzebski JA, Frank GR, Greenleaf JF, Carson PL. Anthropomorphic breast phantoms for assessing ultrasonic imaging system performance and for training ultrasonographers. I. J Clin Ultrasound 1982;10:67–75.
- Madsen EL, Zagzebski JA, Frank GR, Greenleaf JF, Carson PL. Anthropomorphic breast phantoms for assessing ultrasonic imaging system performance and for training ultrasonographers. II. J Clin Ultrasound 1982;10:91–100.
- Madsen EL, Zagzebski JA, Frank GR. An anthropomorphic ultrasound breast phantom containing intermediate-sized scatterers. Ultrasound Med Biol 1982;8:381–392.[CrossRef][Medline]
- Madsen EL, Kelly-Fry E, Frank GR. Anthropomorphic phantoms for assessing systems used in ultrasound imaging of the compressed breast. Ultrasound Med Biol 1988;14(suppl 1):183–201.
- Zagzebski JA, Madsen EL, Frank GR. A teaching phantom for sonographers. J Clin Ultrasound 1991;19:27–38.[Medline]
- Kelly-Fry E, Madsen EL, Frank GR. Use of anthropomorphic breast phantoms for comparing ultrasound breast imagers. Arch Acoust 1991;16:501–511.
- Berg WA, Blume JD, Cormack JB, Mendelson EB, Madsen EL. Lesion detection and characterization in a breast phantom: results of the ACRIN 6666 investigators. Radiology 2006;239(3):693–702.[Abstract/Free Full Text]
- Madsen EL, Zagzebski JA, Banjavic RA, Jutila RE. Tissue-mimicking materials for ultrasound phantoms. Med Phys 1978;5:391–394.[CrossRef][Medline]
- Madsen EL, Zagzebski JA, Frank GR. Oil-in-gelatin dispersions for use as ultrasonically tissue-mimicking materials. Ultrasound Med Biol 1982;8:277–287.[CrossRef][Medline]
- Madsen EL, Dong F, Frank GR, et al. Interlaboratory comparison of ultrasonic backscatter, attenuation, and speed. J Ultrasound Med 1999;18:615–631.[Abstract]
- D'Astous FT, Foster FS. Frequency dependence of ultrasound attenuation and backscatter in breast tissue. Ultrasound Med Biol 1986;12:795–808.[CrossRef][Medline]
- Duck FA. Physical properties of tissue: a comprehensive reference book. London, England: Academic Press, 1990.
This article has been cited by other articles:
|
|
|
|
 
W. A. Berg, J. D. Blume, J. B. Cormack, and E. B. Mendelson
Operator Dependence of Physician-performed Whole-Breast US: Lesion Detection and Characterization.
Radiology,
November 1, 2006;
241(2):
355 - 365.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
