- Email: [email protected]
A fast 3D-imaging technique for performing dynamic Gd-enhanced MRI of breast lesions
Magnetic Resonance Imaging, Vol. 12, No. 4, pp. 545-551, 1994 Copyright 0 1994 Elsevier Science Ltd Printed in the USA. All rights reserved 0730-725X194 56.00 + .CJJ
Pergamon
0730-725X(94)EOOO4-L
l Original Contribution
A FAST 3D-IMAGING TECHNIQUE FOR PERFORMING DYNAMIC Gd-ENHANCED MRI OF BREAST LESIONS WILLIAM H. PERMAN,* ELISAEIETHM. HEIBERG, * JOSEPH GRUNZ, * VIRGINIAM. HERRMANN,? AND CHRISTINAG. JANNEY$ Departmentsof *Radiology,jSurgery,and SPathology,SaintLouis University School of Medicine, St. Louis, MO, USA The characterization of breast lesions by their Gd-enhancement profiles has been proposed as a method for differentiating benign from malignant breast lesions. The limitations of dynamic contrast enhanced 2D imaging of the breast are the low number of slices that can be acquired, and the need to know the location of the lesion a priori to correctly select the noncontiguous 2D slice locations. These problems are exacerbated when multi-focal disease is present but not anticipated. Standard fast 3D gradient-echo imaging has a variable delay between successive acquisitions. We have developed a fast 3D gradient-echo imaging technique for dynamic Gd-DTPA enhanced breast imaging which obtains multiple 3D image sets of 32 contiguous images at 44 s intervals without an interscan delay time. This rapid 3D imaging technique achieves good temporal resolution and reduces patient motion between pre- and postcontrast images while covering a much larger portion of the breast and eliminating the need for a priori knowledge concerning the location of the lesion(s) when performing Gd-enhanced dynamic MR imaging. Keywords: Breast; Neoplasms; MR; Gadolinium; Magnetic resonance (MR), contrast enhancement; Magnetic resonance (MR), pulse sequences.
INTRODUCTION
between normal and malignant breast tissue. Gd-DTPA enhanced MR scans performed using T, weighted spin-
Currently a positive predictive value of 20 to 25% is commonly accepted as reasonable for mammography,’ improvement in specificity by another imaging technique would greatly reduce the number of false positive biopsies performed for clinically or mammographically suspicious lesions. The poor specificity of mammography increases the cost of detecting early cancer, often causes unnecessary patient anxiety, and breast lesion biopsy may cause disfigurement in patients with small breasts. Early clinical evaluations of breast MR imaging were not promising,2-5 partly due to the failure of MR to accurately distinguish malignant from benign breast tissue. In response to this problem several group&” followed the lead from brain tumor imaging and investigated the ability of IV administration of gadopentetate dimeglumine (Gd-DTPA, Berlex Laboratories, Inc., Wayne, New Jersey) to provide enhanced contrast
echo*-” (SE), and gradient-echo’2,‘4*16 imaging with images obtained 5 to 10 min following contrast agent administration were unable to distinguish a malignant mass from a fibroadenoma, mastitis, or proliferative dysplasia solely on the basis of increased signal enhancement. However, when the postGd-DTPA images are acquired within the first 60 to 90 s following injection, both SE6,r0 and GE7 images are reportedly able to discriminate benign from malignant breast tissue in most cases. Initial results indicate that malignant masses enhance rapidly,‘,” reaching 90% of their maximum within 60-90 s following injection.” Fibroadenomas enhance to a greater degree than carcinomas but at a slower rate, exhibiting more enhancement than carcinomas 2-4 min following injection.7~‘0~‘3Therefore, maximum contrast between benign and malignant lesions is demonstrated in images obtained before 90 s following contrast injection. Kaiser” reports a sensi-
RECEIVED 9/13/93; ACCEPTED 12/22/93. Address correspondence to William H. Perman, Ph.D., Department of Radiology, Saint Louis University Medical
Center, 3635 Vista Ave. at Grand Blvd., P.O. Box 15250, St. Louis, MO 63110-0250. 545
546
Magnetic Resonance Imaging 0 Volume 12, Number 4, 1994
tivity of 98.4% for detection of malignant lesions and a specificity of 98.1% for discriminating malignant from benign masses using GE imaging 60 s after contrast injection. One reason for the slightly decreased sensitivity is the inability to localize” some lesions for the accurate placement of the noncontiguous 2D slices used in the pre and postGd-DTPA imaging. The pathologic mechanism which causes increased enhancement of tumors over normal tissue is not known. Since Gd-DTPA escapes from the vascular space but does not cross cell membranes it is known to accumulate in the interstitial space. The high vascularity of tumors10*18and fibroadenomas together with the increased interstitial space and capillary permeability of tumors” probably accounts for the marked increase in MR signal following Gd-DTPA administration. The limitations of dynamic contrast enhanced 2D imaging of the breast are the low number of slices that can be acquired, and the need to know the location of the lesion a priori to correctly select the noncontiguous 2D slice locations. These problems are exacerbated when multi-focal disease is present but not anticipated. Previous attempts at Cd-enhanced 3D gradient-echo imaging required 5 to 10 min to acquire a single data set.12*‘4,16We have developed a fast 3D gradient-echo imaging technique for dynamic Gd-DTPA enhanced breast imaging which obtains multiple 3D image sets of 32 contiguous images at 44 s intervals without an interscan delay time. METHODS
Our contrast enhanced 3D breast MR imaging protocol is as follows: An IV line is placed in the patients arm with sufficient tubing to allow injection of contrast media from outside the bore of the magnet. The patient lies in the prone position on the MR couch with the breast positioned within a 16.5 cm receive-only surface coil (shoulder coil model #412GE-64, Medical Advances, Milwaukee, Wisconsin). The couch is then advanced to the center of a 1.5 T magnet which is part of our General Electric Signa MR scanner (General Electric Medical Systems, Waukesha, Wisconsin) operating in a 4.8x configuration. Axial T,weighted scout scans are acquired at TR = 600 ms, TE = 10 ms, FOV = 320 mm, TH = 5 mm, NEX = 1, and 128 x 256 matrix. Then sagittal T,and fast spin-echo” T2 (ETL = 16) weighted scans are proscribed from the axial scout images and are acquired with TR/TE = 400/l 1 and 3030/90 ms, FOV = 240 mm, TH = 3-4 mm, interslice gap of 1 mm, NEX = 2, and 128 x 256 matrix. Following the T,and T2 weighted scans a selective fast 3D acquisition is graphically prescribed in the sagittal plane to cover the entire breast and chest wall. The fast 3D
spoiled GRASS (F3DSPGR) gradient-echo parameters are TR/TE = 10.6/2.2 ms, FOV = 240 mm, FLIP = 30”, TH = 3-4 mm, 32 z phase-encodings, 128 y phaseencodings, 256 readout points, with a NEX = 1 for an imaging time of 44 s. Without modification, the interscan time between successive 3D data collections on our MR scanner varies between 35 and 50 s for a 32 x 128 x 256 data acquisition. We have modified the F3DSPGR pulse sequence such that multiple 3D data sets (usually 8) are acquired sequentially in time at the same location without an interscan delay following the ‘start scan’ command. Patient movement between the pre and post Gd-DTPA scans is minimized during the multiple 3D acquisition by injecting the Gd-DTPA via previously inserted IV line during the last 10 s of the first 3D acquisition (Fig. 1). Following scanning the images are transferred to an independent computer workstation where difference images as a function of time following Gd-DTPA injection are formed from the pre and postGd images on a pixel by pixel basis. Percent signal enhancement [ lOO(S - So)/ So)] as a function of time following Gd-DTPA injection for lesion, normal breast tissue, and fat are then calculated from the pre and postGd images using regions-of-interest (ROI) locations selected from the difference images. Also, since the dynamic data consists of a 3D data set, lesion shape and relationship to surrounding fat/tissue can be examined by reformatting the images into axial and coronal planes, and by using maximum intensity projections of individual difference 3D data sets. The study population consists of 21 patients presenting with clinical or mammographically suspicious noncystic lesions who are scheduled for surgical breast biopsies. The study patients undergo MRI examination in the few days between the x-ray mammographic diagnosis and the surgical biopsy. Informed consent was obtained prior to the MR examination, and the MRI protocol has been approved by the Saint Louis University Human Subjects Committee. RESULTS
We find that the continuous acquisition of 3D image data before, during and after Gd-DTPA injection allows quantitative analysis of tissue enhancement with neghgible patient motion between successive image data sets. This is demonstrated in Fig. 2 where the first four images of a selected slice from the 8 temporally contiguous 3D acquisitions are shown in the top panel, and the corresponding difference images are shown in the bottom panel. The high fidelity of registration between the baseline 3D data set and the subsequent 3D data sets allows easy visualization of the small (5-6 mm)
Dynamic3D breast imaging0 W.H. PERMANET AL.
547
#l 32 images 4-44
set+
Fig. 1. Depiction of F3DSPGR Gd-enhanced data acquisition. An operator selectable number of 32 image 3D data sets (8 in this example) are acquired in 44 s intervals without an interscan delay time. The IV bolus injection of Gd-DTPA contrast is given during the last 10 s of the first 3D acquisition and is followed by a saline bolus. The phase-encodings are taken in order from -maximum to +maximum encoding gradient amplitude therefore the center of each 3D acquisition corresponds to the zero crossing of the phase-encoding gradient amplitude and is taken as the time point of the data set.
lesion. Any enhancing lesion is easily identified using the difference image method when the subject does not move. Although malignant lesions are usually bright on 7”-weighted images this lesion was not seen on either the fast spin-echo Tz or the standard T, -weighted spin-echo images (not shown). Therefore a priori slice selection for a 2D Gd-enhanced MR study would not have been possible. The percent difference Gdenhancement profile of this lesion is plotted in Fig. 3. Following surgical biopsy this lesion was diagnosed as invasive lobular carcinoma. Note that since fat does not enhance following Gd-DTPA injection, the difference images provide excellent delineation between lesion or breast tissue and fat. We can easily see the “L” shape of the lesion and the relationship of the lesion to the breast anatomy in Fig. 4 where we present the axial maximum intensity projection (MIP) view of this lesion formed from the Iast (t = 4.8 min) differenceimage 3D data. We have studied 3 1 patients with breast lesions using contrast-enhanced MR, 21 of these patients were studied using dynamic volume imaging. Our first 10 patients were studied with a 2D FLASH sequence at one location. Contrast enhanced MR detected a total of 42 lesions, 40 of which have been biopsied. Using the criteria that cancers will exhibit 90% enhancement within the first 90 to 120 s we correctly identified all 8 invasive cancers and 24 of 32 benign lesions giving a sensitivity of 100%) a specificity of 75%) and an accuracy of 80%. The false positive lesions were 3 of 11 fibroadenomas, 4 patients with stromal changes, and 1 patient with lymphadenitis. Though our positive predictive value was only 50%) our negative predictive value was 100%. Three lesions were not seen on T,- and T2-weighted
spin-echo imaging, but were detected with the dynamic contrast enhanced volume study. Two were papillomatous masses that were not detected mammographically, the third was a 7 x 5 x 4 mm nonpalpable invasive lobular carcinoma. Because of our ability to scan the whole breast and axilla dynamically metastatic disease to the axillary lymph nodes and rib metastasis were detected in one patient. DISCUSSION Dynamic Gd-enhancement MR methods utilizing 2D techniques rely upon the visualization of the lesion for the selection of the noncontiguous 2D slices which will be acquired before and after IV Gd-DTPA injection. The number of different 2D slice locations acquired for dynamic Gd-enhanced MR imaging depends upon the selected TR/TE, matrix size, and interscan cycle delays due to reconstruction and downloading of the pulse sequence controller microcode. Kaiser7 initially used a gradient-echo technique with TR/TE = 30/13 ms to acquire a single image every 60 s. As the technology improved he was able to modify his protocol’4 (sacrificing some of the T, weighting) to acquire 5 to 11 slices with TR/TE = 100/5 ms, flip angle = 80”, FOV = 300 mm, and matrix of 256 x 256 pixels. Stack” used a spin-echo technique with TR/TE = 100/28 ms and a 128 x 128 image matrix to acquire 4 images per min at a single slice location. An unmodified F3DSPGR pulse sequence could be implemented to obtain 32 128 x 256 matrix scans in 44 s, however the need to repeatedly initiate the scan process and the variable delays of 35 to 50 s between successive 3D acquisitions significantly degrade the temporal resolution
Magnetic Resonance Imaging 0 Volume 12, Number 4, 1994 B
C
Fig. 2. F3DSPGR images showing the enhancement of a 6mm diameter invasive lobular carcinoma at t = 0 (A,E), 22 (B), 66 (C), and 110 s (D) following IV Gd-DTPA injection. The small lesion is best seen in the difference images F = B - A, G = C - A, and E = D - A.
Dynamic 3D breast imaging 0 W.H. PERMANET AL.
z V
260.0-
& z
220.0-
z 0
180.0-
:c3
140.0-
549
G lOO.O-
(20.0) i
0.0
.
.
.
,
.
.
.
1.0
,
.
.
2.0
.
, 3.0
.
.
.
, 4.0
. 0
Time (min) Fig. 3. The percent difference [ lOO(S - S,)/S,] Gd-DTPA enhancement DTPA injection for a region-of-interest taken in the suspect lesion.
of this unmodified sequence. We have modified the F3DSPGR pulse sequence to remove the large variable interscan delays such that the multiple (operator specified) 32 slice 3D data sets are acquired with only a 50 ms interscan delay between successive 3D data acquisitions. Therefore, the F3DSPGR 3D protocol acquires 3 to 6 times the number of 2D slices as either the Kaiser” or Stack” protocols while maintaining both the T, weighting and contiguous slice locations. The ability to image all of the breast tissue with contiguous slices is important because not all mammographically suspicious lesions are seen with T,and T2 imaging, and there is a 44 to 60% incidence of multicentric breast disease22 in patients with nonpalpable breast cancers detected at mammography. In our study of 21 patients we detected and correctly classified three lesions which were not palpable or detected using mammography. Since the object of the dynamic MR study is to observe the signal changes in breast tissue as a function of time following IV injection of Gd-DTPA, the choice of echo time for the F3DSPGR pulse sequence is de-
plotted as a function
of time following
IV bolus Gd-
termined solely on the basis of maximizing the breast tissue MR signal and minimizing TR and scan time. The MR signal of breast tissue will not be modulated by the fat-water chemical shift therefore we can use the shortest fractional echo time consistent with good image quality. Fat suppression is not needed in our F3DSPGR protocol since images are acquired before and after GdDTPA injection and nonenhancing fat is easily separated from enhancing breast tissue in the difference images (Figs. 2 and 4). recognized the limitation of 2D Harms et a1.‘2,14~‘6 imaging and suggested the use of high spatial resolution (1.4 mm x 0.7 mm x 0.7 mm) 3D gradient-echo imaging i4*16following Gd-DTPA administration to image the entire breast and insure adequate margins for breast-conservation surgery. Since these sequences were applied following Gd-DTPA administration pre and postGd-DTPA difference imaging was not available therefore the authors incorporated fat suppression in the pulse sequences to help differentiate the enhancing lesion from high signal nonenhancing fat.
550
Fig. 4. The axial maximum
Magnetic Resonance Imaging 0 Volume 12, Number 4, 1994
intensity
projection
(MIP) im-
age of the lesion shown in Fig. 2. The MIP image was derived from the difference images created by subtracting the f = 4.8 min 3D data set from the preGd injection data set.
The RODEO (rotating delivery of excitation off resonance) technique16 uses steady-state gradient-echo imaging with fat suppression and acquires a 128 x 256 x 256 3D matrix with TR/TE = 18.5/3.9 ms in 10.1 min, or a 64 x 256 x 256 matrix in 5.1 min. Although the RODEO technique demonstrates high spatial resolution and did not produce any false negatives,16 the long acquisition time prevents the acquisition of Gdenhancement profiles which may be useful for classifying the lesion as benign or malignant as described by Kaiser7 and Stack.” The low positive predictive value of mammography (20-25%)’ and the need for early detection of breast malignancy indicate that better methods are needed to increase the sensitivity and specificity of breast lesion characterization. Since the MR breast exam is expensive, it must show significant improvement in both sensitivity and specificity to justify its use. Increased sensitivity and specificity from the Gd-enhanced MR breast exam may reduce mortality, avoid unnecessary surgical or
needle biopsies, and allow breast conservation therapy (lumpectomy and/or radiotherapy) by detecting all malignant breast lesions and by differentiating benign from malignant lesions on the basis of their enhancement profiles. It is clear from our limited study of 31 patients23 and from the results of others7*10,12,14*16 that most benign lesions do not show significant (~90%) enhancement following IV Gd-DTPA administration. Our preliminary results indicate a two-fold increase in the positive predictive value of dynamic Gd-enhanced MR imaging vs. mammography. We find a sensitivity of lOO%, a specificity of 75%) and an accuracy of 80% when using Gd-enhance MR imaging within the first 90 to 120 s following contrast injection. In addition, three lesions undetected by mammography or palpation were detected and correctly classified. It is also clear that differentiating benign from malignant lesions is not possible in many cases when the postGd MR imaging is performed more than 2 min following Gd injection.7.‘0*14.16On the basis of the results shown in Figs. 2-4 we propose the following protocol for increasing the sensitivity and specificity of the Gd-enhanced MR exam: axial scout scan to localize the breast (1.3 min), and sagittal F3DSPGR scan with 8 temporally contiguous 32 x 128 x 256 3D acquisitions and Gd injection (5.9 min total scan time) for a total MR imaging examination time of 7.2 min. The typical spatial resolution of 3 x 1.4 x 0.9 mm (180 mm FOV) will allow detection of enhancing lesions of size equal to or greater than the voxel dimension. Since the 3 mm slice thickness may degrade lesion detectability, we are also investigating the application of a half-Fourier gradientecho technique24 for dynamic MR imaging as a means of reducing slice thickness to 2 mm while increasing the number of slices from 32 to 64 to minimize partial volume artifacts and insure full breast coverage while maintaining the same 44 s scanning time. In summary, we have developed a fast 3D gradientecho imaging technique for dynamic Gd-DTPA enhanced breast imaging which obtains multiple 3D image sets of 32 contiguous images at 44 s intervals without an interscan delay time. This rapid Gd-enhanced dynamic 3D MR imaging technique achieves good temporal resolution and reduces patient motion between pre- and postcontrast images while covering a much larger portion of the breast than other dynamic scan sequences eliminating the need for a priori knowledge concerning the location of the lesion(s). We are currently investigating whether using this 3D technique will increase the sensitivity and specificity of the MR breast examination while decreasing the overall imaging time to less than 8 min. Acknowledgments-The authors are indebted to Dr. Dennis Parker for his help in modifying the 3D pulse sequence, and Mark Stemm-
uynamlc jy oreasr imagmg w w ler for his help in developing the clinical imaging protocol. The authors would also like to thank the following General Electric Medical Systems personnel: Tom Foo for providing the basic fast 3D SPGR sequence, and Dave Weber, Chuck Lloyd, and Greg Pavlik for their help with understanding and modifying the GE pulse sequences.This work was supported in part by Berlex Laboratories who generously donated the Gd-DTPA contrast used in this study.
12.
REFERENCES 13. 1 Kopans, D.B. Breast Imaging. Philadelphia: JB Lippincott; 1989. 2 Turner, D.A.; Alcorn, F.S.; Adler, Y.T. Nuclear magnetic resonance in the diagnosis of breast cancer. Imaging strategies in MRI. Radiol. Clin. North Am. 26: 673-687; 1988. 3 El Yousef, S.J.; O’Connell, D.M.; Duchesneau, R.H.; Smith, M. J.; Hubay, C.A.; Guyton, S.P. Benign and malignant breast disease: Magnetic resonance and radiofrequency pulse sequences. AJR 145:1-18; 1985. 4. Dash, N.; Lupetin, A.R.; Daffner, R.H.; Deeb, Z.L.; Sefczek, R.J.; Schapiro, R.1. Magnetic resonance imaging in the diagnosis of breast disease. AJR 146: 119-125; 1986. 5. Stelling, C.B.; Powell, D.E.; Mattingly, S.S. Fibroadenomas: Histopathologic and MR imaging features. Radiology 162:399-407;
14.
15. 16.
17. 18.
1987.
6. Heywang, S.H.; Fenzl, G.; Hahn, D.; Krischke, I.; Ed-
19.
maier, M.; Edmaier, M.; Eiermann, W.; Bassermann, R. MR imaging of the breast: Comparison with mammography and ultrasound. J. Comput. Assist. Tomogr. 10:615-20; 1986. I. Kaiser, W.A.; Zeitler, E. MR imaging of the breast: Fast
imaging sequences with and without Gd-DTPA, preliminary observations. Radiology 170:681-86; 1989. 8. Heywang, S.H.; Wolf, A.; Pruss, E.; Hilbertx, T.; Eiermann, W.; Permanetter, W. MR imaging of the breast with Gd-DTPA: Use and limitations. Radiology 171:95103; 1989. 9. Heywang, S.H.; Hilbertz, T.; Beck, R.; Bauer, W.m.;
Eiermann, W.; Permanetter, W. Gd-DTPA enhanced MR imaging of the breast in patients with postoperative scarring and silicon implants. J. Comput. Assist. Tomogr. 14:348-356; 1990. 10. Stack, J.P.; Redmond,O.M.; Codd, M.B.; Dervan,P.A.; Ennis, J.T. Breast disease: Tissue characterization with Gd-DTPA enhancement profiles. Radiology 174:491494; 1990. 11. Rubens,
D.; Totterman,
S.; Chacko,
A.K.; Kothari,
K.;
20
l-l.
k’ERMAN ET AL.
Logan-Young, W.; Szumowski, J.; Simon, J.H.; Zachariah, E. Gadopentetate dimeglumine-enhanced chemical-shift MR imaging of the breast. AJR 157:267270; 1991. Pierce, W.B.; Harms, S.E.; Flamig, D.P.; Griffey, R.H.; Evans, W.P.; Hagans, J.E. Three-dimensional gadolinium-enhanced MR imaging of the breast: pulse sequence with fat suppression and magnetization transfer contrast. Radiology 181:757-763; 1991. Hachiya, J.; Seki, T.; Okada, M.; et al. MR imaging of the breast with Gd-DTPA enhancement: comparison with mammography and ultrasonography. Rad. Med. 9: 232-240; 1991. Harms, S.E.; Flamig, D.P.; Hesley, K.L.; Evans, W.P. Magnetic resonance imaging of the breast. Magn. Reson. Quar. 8:139-155; 1992. Kaiser, W.A. MR mammography (MRM). MedicaMundi 36:168-182; 1991. Harms, S.E.; Flamig, D.P.; Hesley, K.L.; Meiches, M.D.; Jensen, R.A.:Evans, W.P.; Savin0,D.A.; Wells, R.V. MR imaging of the breast with rotating delivery of excitation off resonance: Clinical experience with pathologic correlation. Radiology 187:493-501; 1993. Kaiser, W.A. MRM promises earlier breast cancer diagnosis. Diagn. Zmag. Sept.:88-93; 1992. Cosgrove, D.O.; Bamber, J.C. Color doppler in breast diseases. BRJ 62:659(ab); 1989. Revel, D.; Brasch, R.; Paajanen, H.; et al. Gd-DTPA contrast enhancement and tissue differentiation in MR imaging of experimental breast carcinoma. Radiology 158:339-323; 1986. Hennig, J.; Nauerth, A.; Friedburg, H. RARE imaging: a fast imaging method for clinical MR. Magn. Reson. Med. 3:823-833;
1986.
21 Kaiser, W.A. Dynamic Magnetic Resonance breast imaging using a double breast coil: an important step towards routine examination of the breast. Front. Europ. Rad. 7:39-68; 1990. 22. Holland, R.; Veling, S.H.J.; Mravunac, M.; Hendriks, J.H.C.L. Histologic multifocality of Tis, Tl-2 breast carcinomas: implications for clinical trials of breastconserving surgery. Cancer 56:979-990; 1985. 23. Heiberg, E.M.; Perman, W.H.; Grunz, J.; Herrmann, V.M.; Janney, C.G. Dynamic 3D imaging of the breast. (In preparation.) 24. Perman, W.H.; El-Ghazzawy, 0.; Gado, M.H.; Larson, K.B.; Perlmutter, J.S. A half-Fourier gradient-echo technique for dynamic MR imaging. Mugn. Reson. Imaging 13:357-366; 1993.
Pergamon
0730-725X(94)EOOO4-L
l Original Contribution
A FAST 3D-IMAGING TECHNIQUE FOR PERFORMING DYNAMIC Gd-ENHANCED MRI OF BREAST LESIONS WILLIAM H. PERMAN,* ELISAEIETHM. HEIBERG, * JOSEPH GRUNZ, * VIRGINIAM. HERRMANN,? AND CHRISTINAG. JANNEY$ Departmentsof *Radiology,jSurgery,and SPathology,SaintLouis University School of Medicine, St. Louis, MO, USA The characterization of breast lesions by their Gd-enhancement profiles has been proposed as a method for differentiating benign from malignant breast lesions. The limitations of dynamic contrast enhanced 2D imaging of the breast are the low number of slices that can be acquired, and the need to know the location of the lesion a priori to correctly select the noncontiguous 2D slice locations. These problems are exacerbated when multi-focal disease is present but not anticipated. Standard fast 3D gradient-echo imaging has a variable delay between successive acquisitions. We have developed a fast 3D gradient-echo imaging technique for dynamic Gd-DTPA enhanced breast imaging which obtains multiple 3D image sets of 32 contiguous images at 44 s intervals without an interscan delay time. This rapid 3D imaging technique achieves good temporal resolution and reduces patient motion between pre- and postcontrast images while covering a much larger portion of the breast and eliminating the need for a priori knowledge concerning the location of the lesion(s) when performing Gd-enhanced dynamic MR imaging. Keywords: Breast; Neoplasms; MR; Gadolinium; Magnetic resonance (MR), contrast enhancement; Magnetic resonance (MR), pulse sequences.
INTRODUCTION
between normal and malignant breast tissue. Gd-DTPA enhanced MR scans performed using T, weighted spin-
Currently a positive predictive value of 20 to 25% is commonly accepted as reasonable for mammography,’ improvement in specificity by another imaging technique would greatly reduce the number of false positive biopsies performed for clinically or mammographically suspicious lesions. The poor specificity of mammography increases the cost of detecting early cancer, often causes unnecessary patient anxiety, and breast lesion biopsy may cause disfigurement in patients with small breasts. Early clinical evaluations of breast MR imaging were not promising,2-5 partly due to the failure of MR to accurately distinguish malignant from benign breast tissue. In response to this problem several group&” followed the lead from brain tumor imaging and investigated the ability of IV administration of gadopentetate dimeglumine (Gd-DTPA, Berlex Laboratories, Inc., Wayne, New Jersey) to provide enhanced contrast
echo*-” (SE), and gradient-echo’2,‘4*16 imaging with images obtained 5 to 10 min following contrast agent administration were unable to distinguish a malignant mass from a fibroadenoma, mastitis, or proliferative dysplasia solely on the basis of increased signal enhancement. However, when the postGd-DTPA images are acquired within the first 60 to 90 s following injection, both SE6,r0 and GE7 images are reportedly able to discriminate benign from malignant breast tissue in most cases. Initial results indicate that malignant masses enhance rapidly,‘,” reaching 90% of their maximum within 60-90 s following injection.” Fibroadenomas enhance to a greater degree than carcinomas but at a slower rate, exhibiting more enhancement than carcinomas 2-4 min following injection.7~‘0~‘3Therefore, maximum contrast between benign and malignant lesions is demonstrated in images obtained before 90 s following contrast injection. Kaiser” reports a sensi-
RECEIVED 9/13/93; ACCEPTED 12/22/93. Address correspondence to William H. Perman, Ph.D., Department of Radiology, Saint Louis University Medical
Center, 3635 Vista Ave. at Grand Blvd., P.O. Box 15250, St. Louis, MO 63110-0250. 545
546
Magnetic Resonance Imaging 0 Volume 12, Number 4, 1994
tivity of 98.4% for detection of malignant lesions and a specificity of 98.1% for discriminating malignant from benign masses using GE imaging 60 s after contrast injection. One reason for the slightly decreased sensitivity is the inability to localize” some lesions for the accurate placement of the noncontiguous 2D slices used in the pre and postGd-DTPA imaging. The pathologic mechanism which causes increased enhancement of tumors over normal tissue is not known. Since Gd-DTPA escapes from the vascular space but does not cross cell membranes it is known to accumulate in the interstitial space. The high vascularity of tumors10*18and fibroadenomas together with the increased interstitial space and capillary permeability of tumors” probably accounts for the marked increase in MR signal following Gd-DTPA administration. The limitations of dynamic contrast enhanced 2D imaging of the breast are the low number of slices that can be acquired, and the need to know the location of the lesion a priori to correctly select the noncontiguous 2D slice locations. These problems are exacerbated when multi-focal disease is present but not anticipated. Previous attempts at Cd-enhanced 3D gradient-echo imaging required 5 to 10 min to acquire a single data set.12*‘4,16We have developed a fast 3D gradient-echo imaging technique for dynamic Gd-DTPA enhanced breast imaging which obtains multiple 3D image sets of 32 contiguous images at 44 s intervals without an interscan delay time. METHODS
Our contrast enhanced 3D breast MR imaging protocol is as follows: An IV line is placed in the patients arm with sufficient tubing to allow injection of contrast media from outside the bore of the magnet. The patient lies in the prone position on the MR couch with the breast positioned within a 16.5 cm receive-only surface coil (shoulder coil model #412GE-64, Medical Advances, Milwaukee, Wisconsin). The couch is then advanced to the center of a 1.5 T magnet which is part of our General Electric Signa MR scanner (General Electric Medical Systems, Waukesha, Wisconsin) operating in a 4.8x configuration. Axial T,weighted scout scans are acquired at TR = 600 ms, TE = 10 ms, FOV = 320 mm, TH = 5 mm, NEX = 1, and 128 x 256 matrix. Then sagittal T,and fast spin-echo” T2 (ETL = 16) weighted scans are proscribed from the axial scout images and are acquired with TR/TE = 400/l 1 and 3030/90 ms, FOV = 240 mm, TH = 3-4 mm, interslice gap of 1 mm, NEX = 2, and 128 x 256 matrix. Following the T,and T2 weighted scans a selective fast 3D acquisition is graphically prescribed in the sagittal plane to cover the entire breast and chest wall. The fast 3D
spoiled GRASS (F3DSPGR) gradient-echo parameters are TR/TE = 10.6/2.2 ms, FOV = 240 mm, FLIP = 30”, TH = 3-4 mm, 32 z phase-encodings, 128 y phaseencodings, 256 readout points, with a NEX = 1 for an imaging time of 44 s. Without modification, the interscan time between successive 3D data collections on our MR scanner varies between 35 and 50 s for a 32 x 128 x 256 data acquisition. We have modified the F3DSPGR pulse sequence such that multiple 3D data sets (usually 8) are acquired sequentially in time at the same location without an interscan delay following the ‘start scan’ command. Patient movement between the pre and post Gd-DTPA scans is minimized during the multiple 3D acquisition by injecting the Gd-DTPA via previously inserted IV line during the last 10 s of the first 3D acquisition (Fig. 1). Following scanning the images are transferred to an independent computer workstation where difference images as a function of time following Gd-DTPA injection are formed from the pre and postGd images on a pixel by pixel basis. Percent signal enhancement [ lOO(S - So)/ So)] as a function of time following Gd-DTPA injection for lesion, normal breast tissue, and fat are then calculated from the pre and postGd images using regions-of-interest (ROI) locations selected from the difference images. Also, since the dynamic data consists of a 3D data set, lesion shape and relationship to surrounding fat/tissue can be examined by reformatting the images into axial and coronal planes, and by using maximum intensity projections of individual difference 3D data sets. The study population consists of 21 patients presenting with clinical or mammographically suspicious noncystic lesions who are scheduled for surgical breast biopsies. The study patients undergo MRI examination in the few days between the x-ray mammographic diagnosis and the surgical biopsy. Informed consent was obtained prior to the MR examination, and the MRI protocol has been approved by the Saint Louis University Human Subjects Committee. RESULTS
We find that the continuous acquisition of 3D image data before, during and after Gd-DTPA injection allows quantitative analysis of tissue enhancement with neghgible patient motion between successive image data sets. This is demonstrated in Fig. 2 where the first four images of a selected slice from the 8 temporally contiguous 3D acquisitions are shown in the top panel, and the corresponding difference images are shown in the bottom panel. The high fidelity of registration between the baseline 3D data set and the subsequent 3D data sets allows easy visualization of the small (5-6 mm)
Dynamic3D breast imaging0 W.H. PERMANET AL.
547
#l 32 images 4-44
set+
Fig. 1. Depiction of F3DSPGR Gd-enhanced data acquisition. An operator selectable number of 32 image 3D data sets (8 in this example) are acquired in 44 s intervals without an interscan delay time. The IV bolus injection of Gd-DTPA contrast is given during the last 10 s of the first 3D acquisition and is followed by a saline bolus. The phase-encodings are taken in order from -maximum to +maximum encoding gradient amplitude therefore the center of each 3D acquisition corresponds to the zero crossing of the phase-encoding gradient amplitude and is taken as the time point of the data set.
lesion. Any enhancing lesion is easily identified using the difference image method when the subject does not move. Although malignant lesions are usually bright on 7”-weighted images this lesion was not seen on either the fast spin-echo Tz or the standard T, -weighted spin-echo images (not shown). Therefore a priori slice selection for a 2D Gd-enhanced MR study would not have been possible. The percent difference Gdenhancement profile of this lesion is plotted in Fig. 3. Following surgical biopsy this lesion was diagnosed as invasive lobular carcinoma. Note that since fat does not enhance following Gd-DTPA injection, the difference images provide excellent delineation between lesion or breast tissue and fat. We can easily see the “L” shape of the lesion and the relationship of the lesion to the breast anatomy in Fig. 4 where we present the axial maximum intensity projection (MIP) view of this lesion formed from the Iast (t = 4.8 min) differenceimage 3D data. We have studied 3 1 patients with breast lesions using contrast-enhanced MR, 21 of these patients were studied using dynamic volume imaging. Our first 10 patients were studied with a 2D FLASH sequence at one location. Contrast enhanced MR detected a total of 42 lesions, 40 of which have been biopsied. Using the criteria that cancers will exhibit 90% enhancement within the first 90 to 120 s we correctly identified all 8 invasive cancers and 24 of 32 benign lesions giving a sensitivity of 100%) a specificity of 75%) and an accuracy of 80%. The false positive lesions were 3 of 11 fibroadenomas, 4 patients with stromal changes, and 1 patient with lymphadenitis. Though our positive predictive value was only 50%) our negative predictive value was 100%. Three lesions were not seen on T,- and T2-weighted
spin-echo imaging, but were detected with the dynamic contrast enhanced volume study. Two were papillomatous masses that were not detected mammographically, the third was a 7 x 5 x 4 mm nonpalpable invasive lobular carcinoma. Because of our ability to scan the whole breast and axilla dynamically metastatic disease to the axillary lymph nodes and rib metastasis were detected in one patient. DISCUSSION Dynamic Gd-enhancement MR methods utilizing 2D techniques rely upon the visualization of the lesion for the selection of the noncontiguous 2D slices which will be acquired before and after IV Gd-DTPA injection. The number of different 2D slice locations acquired for dynamic Gd-enhanced MR imaging depends upon the selected TR/TE, matrix size, and interscan cycle delays due to reconstruction and downloading of the pulse sequence controller microcode. Kaiser7 initially used a gradient-echo technique with TR/TE = 30/13 ms to acquire a single image every 60 s. As the technology improved he was able to modify his protocol’4 (sacrificing some of the T, weighting) to acquire 5 to 11 slices with TR/TE = 100/5 ms, flip angle = 80”, FOV = 300 mm, and matrix of 256 x 256 pixels. Stack” used a spin-echo technique with TR/TE = 100/28 ms and a 128 x 128 image matrix to acquire 4 images per min at a single slice location. An unmodified F3DSPGR pulse sequence could be implemented to obtain 32 128 x 256 matrix scans in 44 s, however the need to repeatedly initiate the scan process and the variable delays of 35 to 50 s between successive 3D acquisitions significantly degrade the temporal resolution
Magnetic Resonance Imaging 0 Volume 12, Number 4, 1994 B
C
Fig. 2. F3DSPGR images showing the enhancement of a 6mm diameter invasive lobular carcinoma at t = 0 (A,E), 22 (B), 66 (C), and 110 s (D) following IV Gd-DTPA injection. The small lesion is best seen in the difference images F = B - A, G = C - A, and E = D - A.
Dynamic 3D breast imaging 0 W.H. PERMANET AL.
z V
260.0-
& z
220.0-
z 0
180.0-
:c3
140.0-
549
G lOO.O-
(20.0) i
0.0
.
.
.
,
.
.
.
1.0
,
.
.
2.0
.
, 3.0
.
.
.
, 4.0
. 0
Time (min) Fig. 3. The percent difference [ lOO(S - S,)/S,] Gd-DTPA enhancement DTPA injection for a region-of-interest taken in the suspect lesion.
of this unmodified sequence. We have modified the F3DSPGR pulse sequence to remove the large variable interscan delays such that the multiple (operator specified) 32 slice 3D data sets are acquired with only a 50 ms interscan delay between successive 3D data acquisitions. Therefore, the F3DSPGR 3D protocol acquires 3 to 6 times the number of 2D slices as either the Kaiser” or Stack” protocols while maintaining both the T, weighting and contiguous slice locations. The ability to image all of the breast tissue with contiguous slices is important because not all mammographically suspicious lesions are seen with T,and T2 imaging, and there is a 44 to 60% incidence of multicentric breast disease22 in patients with nonpalpable breast cancers detected at mammography. In our study of 21 patients we detected and correctly classified three lesions which were not palpable or detected using mammography. Since the object of the dynamic MR study is to observe the signal changes in breast tissue as a function of time following IV injection of Gd-DTPA, the choice of echo time for the F3DSPGR pulse sequence is de-
plotted as a function
of time following
IV bolus Gd-
termined solely on the basis of maximizing the breast tissue MR signal and minimizing TR and scan time. The MR signal of breast tissue will not be modulated by the fat-water chemical shift therefore we can use the shortest fractional echo time consistent with good image quality. Fat suppression is not needed in our F3DSPGR protocol since images are acquired before and after GdDTPA injection and nonenhancing fat is easily separated from enhancing breast tissue in the difference images (Figs. 2 and 4). recognized the limitation of 2D Harms et a1.‘2,14~‘6 imaging and suggested the use of high spatial resolution (1.4 mm x 0.7 mm x 0.7 mm) 3D gradient-echo imaging i4*16following Gd-DTPA administration to image the entire breast and insure adequate margins for breast-conservation surgery. Since these sequences were applied following Gd-DTPA administration pre and postGd-DTPA difference imaging was not available therefore the authors incorporated fat suppression in the pulse sequences to help differentiate the enhancing lesion from high signal nonenhancing fat.
550
Fig. 4. The axial maximum
Magnetic Resonance Imaging 0 Volume 12, Number 4, 1994
intensity
projection
(MIP) im-
age of the lesion shown in Fig. 2. The MIP image was derived from the difference images created by subtracting the f = 4.8 min 3D data set from the preGd injection data set.
The RODEO (rotating delivery of excitation off resonance) technique16 uses steady-state gradient-echo imaging with fat suppression and acquires a 128 x 256 x 256 3D matrix with TR/TE = 18.5/3.9 ms in 10.1 min, or a 64 x 256 x 256 matrix in 5.1 min. Although the RODEO technique demonstrates high spatial resolution and did not produce any false negatives,16 the long acquisition time prevents the acquisition of Gdenhancement profiles which may be useful for classifying the lesion as benign or malignant as described by Kaiser7 and Stack.” The low positive predictive value of mammography (20-25%)’ and the need for early detection of breast malignancy indicate that better methods are needed to increase the sensitivity and specificity of breast lesion characterization. Since the MR breast exam is expensive, it must show significant improvement in both sensitivity and specificity to justify its use. Increased sensitivity and specificity from the Gd-enhanced MR breast exam may reduce mortality, avoid unnecessary surgical or
needle biopsies, and allow breast conservation therapy (lumpectomy and/or radiotherapy) by detecting all malignant breast lesions and by differentiating benign from malignant lesions on the basis of their enhancement profiles. It is clear from our limited study of 31 patients23 and from the results of others7*10,12,14*16 that most benign lesions do not show significant (~90%) enhancement following IV Gd-DTPA administration. Our preliminary results indicate a two-fold increase in the positive predictive value of dynamic Gd-enhanced MR imaging vs. mammography. We find a sensitivity of lOO%, a specificity of 75%) and an accuracy of 80% when using Gd-enhance MR imaging within the first 90 to 120 s following contrast injection. In addition, three lesions undetected by mammography or palpation were detected and correctly classified. It is also clear that differentiating benign from malignant lesions is not possible in many cases when the postGd MR imaging is performed more than 2 min following Gd injection.7.‘0*14.16On the basis of the results shown in Figs. 2-4 we propose the following protocol for increasing the sensitivity and specificity of the Gd-enhanced MR exam: axial scout scan to localize the breast (1.3 min), and sagittal F3DSPGR scan with 8 temporally contiguous 32 x 128 x 256 3D acquisitions and Gd injection (5.9 min total scan time) for a total MR imaging examination time of 7.2 min. The typical spatial resolution of 3 x 1.4 x 0.9 mm (180 mm FOV) will allow detection of enhancing lesions of size equal to or greater than the voxel dimension. Since the 3 mm slice thickness may degrade lesion detectability, we are also investigating the application of a half-Fourier gradientecho technique24 for dynamic MR imaging as a means of reducing slice thickness to 2 mm while increasing the number of slices from 32 to 64 to minimize partial volume artifacts and insure full breast coverage while maintaining the same 44 s scanning time. In summary, we have developed a fast 3D gradientecho imaging technique for dynamic Gd-DTPA enhanced breast imaging which obtains multiple 3D image sets of 32 contiguous images at 44 s intervals without an interscan delay time. This rapid Gd-enhanced dynamic 3D MR imaging technique achieves good temporal resolution and reduces patient motion between pre- and postcontrast images while covering a much larger portion of the breast than other dynamic scan sequences eliminating the need for a priori knowledge concerning the location of the lesion(s). We are currently investigating whether using this 3D technique will increase the sensitivity and specificity of the MR breast examination while decreasing the overall imaging time to less than 8 min. Acknowledgments-The authors are indebted to Dr. Dennis Parker for his help in modifying the 3D pulse sequence, and Mark Stemm-
uynamlc jy oreasr imagmg w w ler for his help in developing the clinical imaging protocol. The authors would also like to thank the following General Electric Medical Systems personnel: Tom Foo for providing the basic fast 3D SPGR sequence, and Dave Weber, Chuck Lloyd, and Greg Pavlik for their help with understanding and modifying the GE pulse sequences.This work was supported in part by Berlex Laboratories who generously donated the Gd-DTPA contrast used in this study.
12.
REFERENCES 13. 1 Kopans, D.B. Breast Imaging. Philadelphia: JB Lippincott; 1989. 2 Turner, D.A.; Alcorn, F.S.; Adler, Y.T. Nuclear magnetic resonance in the diagnosis of breast cancer. Imaging strategies in MRI. Radiol. Clin. North Am. 26: 673-687; 1988. 3 El Yousef, S.J.; O’Connell, D.M.; Duchesneau, R.H.; Smith, M. J.; Hubay, C.A.; Guyton, S.P. Benign and malignant breast disease: Magnetic resonance and radiofrequency pulse sequences. AJR 145:1-18; 1985. 4. Dash, N.; Lupetin, A.R.; Daffner, R.H.; Deeb, Z.L.; Sefczek, R.J.; Schapiro, R.1. Magnetic resonance imaging in the diagnosis of breast disease. AJR 146: 119-125; 1986. 5. Stelling, C.B.; Powell, D.E.; Mattingly, S.S. Fibroadenomas: Histopathologic and MR imaging features. Radiology 162:399-407;
14.
15. 16.
17. 18.
1987.
6. Heywang, S.H.; Fenzl, G.; Hahn, D.; Krischke, I.; Ed-
19.
maier, M.; Edmaier, M.; Eiermann, W.; Bassermann, R. MR imaging of the breast: Comparison with mammography and ultrasound. J. Comput. Assist. Tomogr. 10:615-20; 1986. I. Kaiser, W.A.; Zeitler, E. MR imaging of the breast: Fast
imaging sequences with and without Gd-DTPA, preliminary observations. Radiology 170:681-86; 1989. 8. Heywang, S.H.; Wolf, A.; Pruss, E.; Hilbertx, T.; Eiermann, W.; Permanetter, W. MR imaging of the breast with Gd-DTPA: Use and limitations. Radiology 171:95103; 1989. 9. Heywang, S.H.; Hilbertz, T.; Beck, R.; Bauer, W.m.;
Eiermann, W.; Permanetter, W. Gd-DTPA enhanced MR imaging of the breast in patients with postoperative scarring and silicon implants. J. Comput. Assist. Tomogr. 14:348-356; 1990. 10. Stack, J.P.; Redmond,O.M.; Codd, M.B.; Dervan,P.A.; Ennis, J.T. Breast disease: Tissue characterization with Gd-DTPA enhancement profiles. Radiology 174:491494; 1990. 11. Rubens,
D.; Totterman,
S.; Chacko,
A.K.; Kothari,
K.;
20
l-l.
k’ERMAN ET AL.
Logan-Young, W.; Szumowski, J.; Simon, J.H.; Zachariah, E. Gadopentetate dimeglumine-enhanced chemical-shift MR imaging of the breast. AJR 157:267270; 1991. Pierce, W.B.; Harms, S.E.; Flamig, D.P.; Griffey, R.H.; Evans, W.P.; Hagans, J.E. Three-dimensional gadolinium-enhanced MR imaging of the breast: pulse sequence with fat suppression and magnetization transfer contrast. Radiology 181:757-763; 1991. Hachiya, J.; Seki, T.; Okada, M.; et al. MR imaging of the breast with Gd-DTPA enhancement: comparison with mammography and ultrasonography. Rad. Med. 9: 232-240; 1991. Harms, S.E.; Flamig, D.P.; Hesley, K.L.; Evans, W.P. Magnetic resonance imaging of the breast. Magn. Reson. Quar. 8:139-155; 1992. Kaiser, W.A. MR mammography (MRM). MedicaMundi 36:168-182; 1991. Harms, S.E.; Flamig, D.P.; Hesley, K.L.; Meiches, M.D.; Jensen, R.A.:Evans, W.P.; Savin0,D.A.; Wells, R.V. MR imaging of the breast with rotating delivery of excitation off resonance: Clinical experience with pathologic correlation. Radiology 187:493-501; 1993. Kaiser, W.A. MRM promises earlier breast cancer diagnosis. Diagn. Zmag. Sept.:88-93; 1992. Cosgrove, D.O.; Bamber, J.C. Color doppler in breast diseases. BRJ 62:659(ab); 1989. Revel, D.; Brasch, R.; Paajanen, H.; et al. Gd-DTPA contrast enhancement and tissue differentiation in MR imaging of experimental breast carcinoma. Radiology 158:339-323; 1986. Hennig, J.; Nauerth, A.; Friedburg, H. RARE imaging: a fast imaging method for clinical MR. Magn. Reson. Med. 3:823-833;
1986.
21 Kaiser, W.A. Dynamic Magnetic Resonance breast imaging using a double breast coil: an important step towards routine examination of the breast. Front. Europ. Rad. 7:39-68; 1990. 22. Holland, R.; Veling, S.H.J.; Mravunac, M.; Hendriks, J.H.C.L. Histologic multifocality of Tis, Tl-2 breast carcinomas: implications for clinical trials of breastconserving surgery. Cancer 56:979-990; 1985. 23. Heiberg, E.M.; Perman, W.H.; Grunz, J.; Herrmann, V.M.; Janney, C.G. Dynamic 3D imaging of the breast. (In preparation.) 24. Perman, W.H.; El-Ghazzawy, 0.; Gado, M.H.; Larson, K.B.; Perlmutter, J.S. A half-Fourier gradient-echo technique for dynamic MR imaging. Mugn. Reson. Imaging 13:357-366; 1993.