May 5, 2009 - Erich Salomonowitz. Received: 3 ... e-mail: erich.salomonowitz@stpoelten. lknoe.at. Tel. ..... land R, Uder M, Kiefer B, Bautz W,. Janka R (2007) ...
Eur Radiol (2009) 19: 2349–2356 DOI 10.1007/s00330-009-1426-2
BREAST
Andreas Stadlbauer Reinhard Bernt Stephan Gruber Wolfgang Bogner Katja Pinker Wilma van der Riet Jörg Haller Erich Salomonowitz
Diffusion-weighted MR imaging with background body signal suppression (DWIBS) for the diagnosis of malignant and benign breast lesions
Received: 3 November 2008 Accepted: 19 March 2009 Published online: 5 May 2009 # European Society of Radiology 2009
W. van der Riet European MRI Consultancy (EMRIC), Strasbourg, France
A. Stadlbauer . E. Salomonowitz MR Physics Group, Department of Radiology, Landesklinikum St. Poelten, St. Poelten, Austria A. Stadlbauer Department of Neurosurgery, University of Erlangen-Nuremberg, Erlangen, Germany R. Bernt . J. Haller Department of Radiology, Hanusch Krankenhaus, Vienna, Austria S. Gruber . W. Bogner . K. Pinker Department of Radiology, Medical University, Vienna, Austria
E. Salomonowitz (*) Department of Radiology, Landesklinikum St. Poelten, Propst Fuehrer Strasse 4, 3100 St. Poelten, Austria e-mail: erich.salomonowitz@stpoelten. lknoe.at Tel.: +43-2742-30018009 Fax: +43-2742-30018019
Abstract The purpose of this study was to evaluate the efficacy of diffusion-weighted MR imaging with background body signal suppression (DWIBS) and a conventional DWI (cDWI) sequence for the detection of breast lesions. Fifty consecutive patients with suspected breast lesions underwent DWIBS and cDWI at 1.5 T. The routine protocol consisted of a short TI inversion recovery
Introduction Diffusion-weighted magnetic resonance imaging (DWI) provides information about the local microstructural characteristics of the diffusivity of water molecules in tissues. The major role of DWI in clinical routine is in the early detection of cerebral ischaemia, but changes in tissue water diffusion properties can be helpful for the detection and characterisation of pathological processes, including cancer, in any part of the body. For many years, owing to technical difficulties, the acceptance of the application of DWI in clinical routine has been limited to examinations of the brain. In recent years, however, new developments in
(STIR) sequence and a dynamic contrast-enhanced T1-weighted sequence. Apparent diffusion coefficient (ADC) and exponential ADC (eADC) values of the lesions were calculated. Receiver operating characteristic (ROC) analyses and qualitative evaluation of lesion detectability and conspicuity were performed. Thirty-six lesions were detected in 30 patients by using the routine protocol. DWIBS detected 34 lesions (94%) and cDWI detected 26 lesions (72%). The conspicuity of fibroadenomas was significantly (P= 0.007) better for DWIBS. ADC and eADC values of tumour were significantly different between DWIBS and cDWI. DWIBS is superior to cDWI in the visualization of malignant and benign lesions in the breast. Keywords Diffusion-weighted MR imaging . Breast cancer . Accuracy
imaging techniques (e.g. parallel imaging) and hardware (e.g. stronger gradient systems and multichannel coils) have helped to overcome several complications (e.g. susceptibility and respiratory motion artefacts) that occurred when DWI sequences were used for examination of parts of the body other than the brain. Hence, in the last few years, the potential of DWI for clinical diagnostics, especially for tumour identification, has been shown for several organs, e.g. liver, kidneys, pancreas, prostate, breast etc. [1, 2], and whole body application [3–6]. Takahara et al. developed and introduced, in 2004, a DWI technique that used a short TI inversion recovery (STIR)–echo planar imaging (EPI) sequence and free
2350
breathing to screen for malignancies in the whole body. This technique was named DWIBS, which stands for diffusion-weighted whole body imaging with background body signal suppression [7]. The usability of DWIBS for the detection of thoracic and abdominal lesions has been demonstrated in a few reports; however, these studies had heterogeneous patient populations and low patient numbers [8–10]. To establish the value of DWIBS and to develop criteria for differentiation between malignant and benign lesions, Kwee et al. recommended studies with a larger sample size of lesions with specific pathological features [11]. The purpose of this study was to evaluate the usefulness of DWIBS compared with a conventional DWI sequence and to define criteria for detecting the malignancy of breast lesions by using DWIBS.
Materials and methods Patients The study was approved by our institutional review board and written, informed consent was obtained from all patients. A total of 50 consecutive patients (age range, 22– 75 years; mean age ± standard deviation (SD), 50.6± 11.9 years) with suspected breast lesions detected by either mammography or ultrasonography were included in this study. Thirty-six lesions were detected in 30 patients (age range, 22–75 years; mean age ± SD, 49.8±11.9 years) by using the routine MR protocol. In patients with suspicious lesions, histology was obtained after the MRI examination through sonographically guided core needle biopsy or stereotactic vacuum-assisted biopsy. The histological evaluation revealed: nine tumours (invasive ductal carcinoma or ductal carcinoma in situ), 12 fibroadenomas, five cysts, seven lymph nodes and three postoperative scars. In premenopausal women, the MR examinations were scheduled for the second week of the menstrual cycle.
336×600, 50 slices with 3-mm slice thickness and without gaps, 3 averages, turbo factor 59, SENSE factor 1.7, resulting in a voxel size of 0.75×0.75× 3.0 mm3); and (c) A three-dimensional (3D) dynamic, contrast-enhanced (CE) T1-weighted gradient-echo sequence (TR/TE= 4.4/2.0 ms, FOV=250×450×150 mm (AP×RL×FH), matrix 168×300, 100 slices with 1.5-mm slice thickness, turbo factor 50, SENSE factor 1.6, 6 dynamic acquisitions, resulting in 1.5-mm3 isotropic voxels, a dynamic data acquisition time of 1 min 30 s, and a total sequence duration of 9 min) with spectral attenuated inversion recovery (SPAIR) fat suppression. A gadolinium chelate contrast medium (Dotarem, Guerbet) was injected at a dose of 0.1 mmol/kg body weight. DWIBS was performed in the transversal orientation using the following parameters: TR/TE/TI= 6,900/65/180 ms; FOV=250×350 mm; matrix 80×112; 50 slices with 3-mm slice thickness and without gaps; b0 =0 s/ mm2; b1 =1,000 s/mm2; 4 averages; EPI factor 67; SENSE factor 1.8; resulting in a voxel size of 3.1×3.1×3.0 mm3 and a data acquisition time of 2 min 4 s. A conventional DWI (cDWI) sequence was also performed using spectral presaturation with inversion recovery (SPIR) fat-suppressed and a TR/TE=6,100/65 ms, a FOV=250×350 mm, a matrix of 80×112, 50 slices with 3-mm slice thickness and without gap, b0 =0 s/mm2, b1 =1,000 s/mm2, 4 averages, an EPI factor of 67, a SENSE factor of 1.8, resulting in a voxel size of 3.1×3.1×3.0 mm3 and a data acquisition time of 1 min 50 s. Both diffusion-weighted MR sequences were performed before the dynamic CE-T1w sequence. Data processing
All MR examinations were performed on a 1.5-T MR system (Achieva, Philips Medical Systems, Best, the Netherlands) using a four-channel breast coil. The routine MRI protocol consisted of:
Qualitative evaluation of lesion detectability and conspicuity with DWBIS and cDWI was performed by two experienced radiologists in consensus. A scale of 0–3 (0= not detectable, 1=poor, 2=acceptable and 3=good) was used for classification. Apparent diffusion coefficient (ADC) and exponential ADC (eADC) values of the lesions and normal breast parenchyma were calculated for both sequences with the following equation: Sb S0 ¼ exp½ðb ADC Þ ¼ eADC; (1)
(a) A transversal short TI inversion recovery (STIR) turbospin-echo (TSE) sequence (TR/TE/TI= 3,800/60/165 ms, field of view (FOV)=250×450 mm (AP×RL), matrix 168×300, 50 slices with 3-mm slice thickness and without gaps, 3 averages, turbo factor 23, resulting in a voxel size of 1.5×1.5×3.0 mm3); (b) A transverse T2-weighted TSE (TR/TE= 6,300/130 ms, FOV=250×450 mm (AP×RL), matrix
b stands for the b value of the DWI sequence (i.e. b1 = 1,000 s/mm2 in our study), and Sb and S0 are the signal intensities on the diffusion-weighted image and the reference image, respectively. The signal intensities were determined by drawing two-dimensional (2D) regions of interest (ROIs) on the images without and with diffusion weighting (b0 and b1), using a DICOM image viewer provided by the vendor of the
MR imaging methods
2351
MR system (DICOM Viewer R2.4, Philips Medical Systems, Best, the Netherlands). For lesions, the ROIs were adapted to cover the whole lesion at its maximum extent on the transversal images. The ROIs in the normal breast parenchyma were defined 1.5 cm behind the mammilla, were ellipseshaped and had a diameter of 40×20 mm. The mean signal intensities in the ROIs were automatically calculated by the software. ADC and eADC values were calculated using Eq. (1) and EXCEL software (Microsoft Office 2007, Microsoft, Redmond, USA). For three-dimensional display of the diffusion-weighted data from the DWIBS and cDWI sequences, maximum intensity projection (MIP) was employed, followed by gray-scale inversion. An MIP reconstruction was also obtained for the STIR sequence and the last dynamic scan of the CE-T1w sequence.
using a Wilcoxon signed rank test. Comparisons of ADC and eADC values between the different types of lesions were performed using the Mann–Whitney U test. For comparisons of ADC and eADC values of lesions and normal breast parenchyma measured with DWIBS and cDWI we used a Wilcoxon signed rank test. Probability values of less than 0.05 were considered significant. Receiver operating characteristic (ROC) analysis [12] was performed to determine suitable cutoff ADC and eADC values for differentiating between individual lesion types and between malignant and benign lesions. The area under the curve (AUC) was used as an index for evaluating the inherent capacity of the diffusion-weighted sequences to differentiate between lesion types.
Results Statistical methods Lesion detectability and conspicuity Data were analysed using statistical software (SPSS, version 14; SPSS, Chicago, USA). The results for the scaling of detectability and conspicuity were compared
Fig. 1 Transverse a DWIBS, b cDWI and c CE-T1w image of the last dynamic measurement from a 22-year-old patient with a fibroadenoma. The lesion is distinguishable on DWIBS, but not on cDWI images. Transverse d DWIBS, e cDWI and f CE-T1w image of the last dynamic measurement from a 57-year-old patient with a
DWIBS detected 34 (94%) of 36 lesions in the 30 patients examined with the routine protocol. The lesions missed
multicentric carcinoma (invasive ductal carcinoma). The lesions are distinguishable on both DWIBS and cDWI; however, the inner structure of the anterior lesion is depicted on DWIBS only and comparable to the findings on the CE-T1w image
2352
were two fibroadenomas. However, only 26 lesions (72%) were detected by cDWI. The lesions missed were one tumour (invasive ductal carcinoma), five fibroadenomas, one cyst, one benign lymph node and two scars. The conspicuity of tumours was good in all nine cases on DWIBS images, whereas cDWI showed good conspicuity in seven cases (78%) and acceptable conspicuity in one case (11%). For fibroadenomas we found good conspicuity on DWIBS images in six cases (50%), acceptable conspicuity in three cases (25%) and poor conspicuity in one case (8%). The conspicuity of fibroadenomas on cDWI data was not rated good for any patients, acceptable in one (8%) and poor in six patients (50%). The conspicuity on DWIBS of the other lesion types was as follows: cysts, all Fig. 2 Sagittal and transverse MIP reconstructions of CE-T1w data from the last dynamic measurement (a, b) and STIR data (c, d). Sagittal and transversal MIP reconstructions and inversion of gray-scale for the DWIBS (e, f) and cDWI (g, h) data from a 50-year-old patient with a fibroadenoma in the left breast. The sagittal MIP reconstructions were performed only for the left breast. Detectability and conspicuity are superior on DWIBS compared with cDWI
(100%) good; lymph nodes, four (57%) good and three (43%) acceptable; and scars, two (67%) good and one (33%) poor. We found that the conspicuity of these lesions on cDWI was: cysts, two (40%) good, one (20%) acceptable and one (20%) poor; lymph nodes, three (43%) good and three (43%) acceptable; and scars, one (33%) poor. However, only the conspicuity of fibroadenomas was significantly (P>=0.007) different between DWIBS (mean scale, 2.1) and cDWI (mean scale, 0.7). The mean scales were: tumours, 3.0 for DWIBS and 2.6 for cDWI; cysts, 3.0 for DWIBS and 1.8 for cDWI; lymph nodes, 2.6 for DWIBS and 2.0 for cDWI; and scar, 2.3 for DWIBS and 0.3 for cDWI. Figure 1 illustrates the differences in lesion detectability and conspicuity between
2353
Fig. 3 Boxplots of ADC values for normal breast parenchyma (“normal”), cyst, fibroadenoma (FA), lymph node, scar and tumour calculated from a DWIBS and b cDWI data. The horizontal lines are the medians; the ends of the boxes show the lower and upper quartiles (25th and 75th percentiles)
DWIBS and cDWI. MIP reconstructions of STIR and CET1w images of the last dynamic measurement and of DWIBS and cDWI data after inversion of gray-scale in a patient with a fibroadenoma are depicted in Fig. 2. ADC and eADC values The box plots in Fig. 3 provide overviews of the ADC values for normal breast tissue and lesions measured with DWIBS (Fig. 3a) and cDWI (Fig. 3b). The ADC values of tumours were significantly different from those of cysts (DWIBS, P=0.001; cDWI, P=0.001), fibroadenoma (DWIBS, P
