| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Innovation |
1 Department of Medical Imaging, Sunnybrook Health Sciences Centre, University
of Toronto, 2075 Bayview Ave., MG166, Toronto, ON, M4N 3M5, Canada.
2 Sentinelle Medical, Toronto, ON, Canada.
3 Departments of Imaging Research and Medical Biophysics, Sunnybrook Health
Sciences Centre, University of Toronto, Toronto, ON, Canada.
Received December 4, 2007;
accepted after revision April 20, 2008.
Address correspondence to P. A. Causer
(Petrina.Causer{at}sunnybrook.ca).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. In 10 patients with 13 lesions, the system was found to be an accurate means for targeting sonography to MRI of the same breast lesions.
Keywords: breast cancer • MRI • sonography
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
However, at evaluation after MRI–sonography, the target lesion may not be visualized. LaTrenta et al. [4] reported that only 21 of 93 (23%) mammographically occult MRI-detected lesions sent to sonography were sonographically visible. Even if a lesion is identified after MRI–sonography examination, it is often unclear whether that lesion corresponds to the MRI-detected lesion in question. MRI-guided breast biopsy is used with increasing frequency, particularly for biopsy of lesions that are visible only on MRI but are occult at mammography and conventional sonography. However, the limitations of MRI-guided procedures include the high costs of MRI-compatible disposables and time-consuming use of limited MRI resources.
Because the sampling is not performed under real-time visualization, as with sonography, and because it is not currently practical to image the sample with MRI to confirm retrieval of the targeted lesion as with stereotactic biopsy of calcifications, biopsy accuracy can be an issue in certain cases. As many as 4% of MRI-guided biopsies may miss the intended target [5].
To ensure that a suspicious lesion visible only on MRI is accurately sampled using sonographic guidance, previous in vitro work documented the accuracy of a coregistration system using sonography to locate an MRI-visible lesion in a breast phantom [6]. The purpose of this pilot study was to determine the accuracy of the same MRI–sonography coregistration system in vivo for showing breast lesions visible on MRI and sonography.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
From the first cohort of women with breast cysts, lesions were diagnosed as simple cysts on the basis of the sonographic features and confirmed on MRI. The three MRI-detected masses were classified as BI-RADS category 4 or 5 and were diagnosed by MRI-guided biopsy (n = 1) or sonography-guided biopsy (n = 2).
Breast MRI and Sonographic Coregistration Equipment
Both breast MRI and sonography were performed with the patient in the prone
position using a system designed at our institution that featured a redesign
of the MRI bed and coil system. The stretcher and supporting table design
allowed open access to the medial and lateral breast to facilitate MRI-guided
biopsy as previously described
[7] or sonographic
coregistration (Fig. 1A,
1B,
1C,
1D). The mobile stretcher was
designed for use with a 1.5-T magnet (Signa, GE Healthcare).
|
|
|
|
The breast was immobilized between medial and lateral compression frames containing a large open aperture, allowing the ultrasound transducer medial or lateral access to the breast through a thin sonography-compatible membrane (Fig. 1B). Fixation points on the frame allowed attachment and exchange of the MR receiver coils [8, 9] for the ultrasound transducer positioning stage (Fig. 1C). MRI-visible fiducial markers were set in known locations within the frame.
The stage positioned the transducer in three dimensions relative to the frame and the patient including x or superoinferior, y or anteroposterior, and z or depth from skin and with two 35° angles of freedom (elevation and azimuth), totaling 5 df. The 5 df of the stage position were calculated on the basis of the MRI coordinates of the lesion relative to the fiducial markers using a computer program.
MRI–Sonography Coregistration Procedure
A 1.5-T magnet (Signa, GE Healthcare) was used. Phased-array bilateral
breast coils of our own institutional design were used. After multiplanar
localizer images were obtained, T1-weighted spin-echo imaging for fiducial
markers was performed with the following parameters: TR/TE, 350/14; matrix,
128 x 256; and field of view, 25 cm. Coronal T1-weighted 2D fast spoiled
gradient-recalled echo (FSPGR) (150/4.2; flip angle, 50°; matrix, 256
x 256; and field of view, 20 cm) and sagittal unenhanced T2-weighted
fast spin-echo fat-suppressed (3,000/102; matrix, 128 x 256; and field
of view, 20 cm) sequences were then performed.
Additional imaging for masses included administration of 0.1 mmol/kg gadodiamide (Omniscan, GE Healthcare) by rapid injection. Three series of unenhanced sagittal T1-weighted 2D FSPGR fat-suppressed images followed by 11 series of contrast-enhanced images were performed with the following parameters: 150/4.2; flip angle, 50°; matrix, 128 x 256; field of view, 20 cm; and slice thickness, 3 mm. The scanning time for each series was 20 seconds. A sagittal T1-weighted 3D FSPGR fat-suppressed image was then obtain ed with the following parameters: 50/4.2; flip angle, 50°; matrix, 256 x 512; field of view, 20 cm; and slice thickness, 1 mm. The scanning time was 6 minutes 50 seconds. A final set of three dyna mic series of delayed sagittal T1-weighted 2D FSPGR fat-suppressed images were then performed.
Once the target lesion was identified at the MR console, the lesion coordinates were obtained from the T2-weighted images for cysts or from the first contrast-enhanced T1-weighted 2D FSPGR fat-suppressed image in which a solid mass was visualized. The location of the lesion was determined from its MRI coordinates relative to the fiducial markers. A computer program was used to calculate the 5 df of the ultrasound transducer stage, which would allow the ultrasound beam to intercept the lesion of interest, displaying the lesion centered in the sonographic field of view. The calculation was based on lesion and fiducial marker coordinates.
|
|
|
Sonography was performed using a linear 5-12–MHz transducer (HDI 5000, Philips Healthcare). The transducer was placed in the stage, based on the MRI-predicted lesion location, according to the coordinates calculated with respect to the patient. The x (superoinferior), y (anteroposterior), and z (depth from skin) co ordinates were dialed in as were the elevation and azimuth angles (Fig. 1C).
The coronal T1-weighted 2D FSPGR images served as a reference to correlate breast parenchymal patterns between MR images and sonograms and to calculate speed-of-sound variations through fat.
Data and Statistical Analyses
Lesion size in the x dimension was measured from the MR images and
sonograms. The mean lesion size based on MRI and sonography was correlated
using linear regression analysis.
In addition to lesion size and morphology, surrounding breast parenchymal interfaces and features were used to confirm that the same lesion was identified on both techniques (Figs. 2A, 2B, 2C and 3A, 3B, 3C). Lesion pathology was determined on the basis of imaging features diagnostic of cysts or histopathology for masses.
|
|
|
The true sonographic position error measurements were calculated in the x and z planes from the sonogram by determining the difference in the center of the displayed lesion coordinates and initial lesion center coordinates calculated on the basis of the MRI. This measurement was per formed prospectively by placing a cursor on the sonography monitor at the MRI-predicted co ordinates and measuring the distance to the sonography lesion center coordinates for both the x and z planes (Fig. 3B). If the target lesion was not immediately displayed, based on the calcu lated stage coordinates, the stage and ultrasound transducer were repositioned in the y-plane until the targeted lesion was displayed. The error measurement for the y-plane was then calculated on the basis of the difference in the initially predicted and the final repositioned stage positions.
Because the speed of sound differs in fat compared with soft tissue, there is an anticipated error in the z-plane. To correct for this error, we retro spectively measured the distance of fat between the skin surface and the lesion on the coronal T1-weighted MR image (Fig. 3C). On the basis of this measurement, a correction was calculated defined as the product of the sonographic propagation distance through fat and the ratio of the speed of sound through fat divided by the speed of sound in soft (fibroglandular) tissue, as previously described [6]. The initial z-plane error measurement was then recalculated based on this correction.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The mean error measurements in the three planes were as follows: x-plane, 2.5 mm (range, 0.9–6.3 mm); y-plane, 1.1 mm (range, 0–4 mm); and z-plane, –2.6 mm (range, –0.9 to 5.3 mm). The mean sonographic speed of sound through fat correction value was 1 mm (range, 0.5–2.2 mm). After applying the correction value to the initially calculated z-plane error measurement, the z-plane error decreased to –1.7 mm (range, –0.04 to 4.2 mm).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The good correlation of the MRI and sonographic lesion size can be explained by the types of lesions included: focal masses and cysts. The reproducible size between the imaging studies was anticipated and was one factor used to ensure the same lesion was being documented. Breast parenchymal landmarks, including fat–fibroglandular interfaces visible on both MRI and sonography, are another useful means to ensure that the same MRI target area is shown with sonography.
Anticipated future clinical applications using this MRI–sonography coregistration system include roles in both diagnostic work-up and biopsy guidance of MRI-detected lesions. One such application is to ensure that a sonographic finding corresponds to an MRI lesion in question when there is doubt using traditional supine sonography. Once a sonographic correlative lesion is confidently identified, it can be appropriately managed using sonographic guidance, including follow-up or biopsy.
An anticipated future use of this system is sonography-guided biopsy of an MRI-detected lesion. Even though MRI-guided percutaneous vacuum-assisted biopsy has become the standard of care for biopsy of suspicious lesions visible only on MRI or lesions for which a sonographic correlative lesion cannot be found with confidence, there is a small but real potential for sampling error. Prior studies that included relatively small numbers have found MRI-guided vacuum-assisted biopsy to be very accurate, with false-negative rates of 6% reported in an in vitro phantom study and ranging from 2% to 4% in vivo [5, 10–12]. However, in reality, imaging–histopathologic correlation and MRI follow-up are most often the only methods from which to infer accurate sampling of benign lesions. Often, on the basis of MRI, there is no direct evidence that the lesion has been accurately biopsied because of the small size or nature of the lesion and parenchymal enhancement, hematoma, and air introduced at the time of biopsy obscuring the target lesion. A previous report using a prototype of the MRI–sonography coregistration system used in this study for in vitro biopsy using a breast phantom reported a higher accuracy using sonography as the method of needle guidance for biopsy compared with MRI alone [6].
In addition to potentially improving sampling accuracy, the overall cost of the biopsy could be reduced by eliminating the need for expensive MR-compatible percutaneous devices. Also, because the biopsy would be performed outside of the magnet, efficiency of magnet use could be improved by the continued imaging of other patients during the biopsy.
One limitation of this study is the inability of other institutions to reproduce the study or results because this system is currently in development and not yet commercially available. However, this study was undertaken to confirm system accuracy before proceeding with further development. The small number of lesions included in the study is a further study limitation for evaluating both targeting error and the physical constraints of the system for lesion evaluation.
In conclusion, our system offers an accurate means for targeting sonography to MRI of the same breast lesions. Currently, this system enables confident identification of a sonographic correlate to an MRI lesion for lesions visible on sonography. As the targeting accuracy of the system is confirmed, further system development can proceed, including a sonographically guided biopsy mechanism using this system for lesions visible only on MRI.
Acknowledgments
We thank Joan Glazier for her valuable contribution in patient
recruitment.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Nakano, M. Yoshida, K. Fujii, K. Yorozuya, Y. Mouri, J. Kousaka, T. Fukutomi, J. Kimura, T. Ishiguchi, K. Ohno, et al. Fusion of MRI and Sonography Image for Breast Cancer Evaluation Using Real-time Virtual Sonography with Magnetic Navigation: First Experience Jpn. J. Clin. Oncol., September 1, 2009; 39(9): 552 - 559. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
