A systematic comparison of two pulse sequences for edema assessment in MR-mammography

Materials and methods 321 consecutive patients from a 22 month period were included in this investigation. Further selection criteria were histopathological verification of enhancing lesions and absence of presurgical chemotherapy or biopsy, resulting in 108 malignant and 107 benign lesions. All underwent MRM according to international guidelines including a non-fatsat T2w-TSE sequence (TR/TE: 8900/207 ms) and a short tau inversion recovery fatsat sequence (STIR, TR/TE: 8420/70 ms). All images were acquired in the same orientation (axial) and slice thickness. Two experienced radiologists in consensus rated presence of perifocal edema according to an ordinal scale: 0 = not present, 1 = little, 2 = intermediate, and 3 = distinct. Data analysis was performed using crosstabs and Visual Grading Characteristics (VGC) analysis. Results Overall sensitivity/specificity was calculated with 53.7%/94.4% (T2w-TSE) and 52.8%/95.3% (STIR). VGC revealed an area under the VGC curve of 0.502 (standard error 0.026), P = 0.814. Conclusion Perifocal edema is a specific feature of malignancy with moderate sensitivity. VGC analysis did not reveal significant differences between both pulse sequences analysed. Consequently, both T2w-TSE and STIR images are suitable for assessment of perifocal edema. Keywords MR-mammography Breast MRI Sensitivity and specificity Edema BI-RADS Breast cancer 1 Introduction Edema describes a pathological increase of fluid content in tissues. This condition may be caused by cardiac insufficiency, inflammation and neoplasms. The presence of edema in both latter conditions is attributed to an increased vascular permeability; possibly also to proteolytic activity during extracellular matrix degradation [1–4] . In magnetic resonance imaging (MRI), the assessment of both presence and location of edema provides important information for detection and differentiation of neoplastic from inflammatory lesions [5] . Furthermore, an association with prognostic factors such as tumor grading (i.e. aggressiveness) has been described [5] . Edema can be visualized by T2-weighted (T2w) sequences based on transverse relaxivity differences. Due to the long T2 relaxation time of water, edema appears bright on T2-weighted images. To increase contrast between fat and water in T2w imaging, long echo times or fat saturation techniques are applied [5,6] . The advantage of unsaturated T2w imaging is superior Signal-to-Noise-Ratio (SNR) and high spatial resolution. The fatsat technique mostly applied to T2w imaging is an inversion recovery technique with short inversion times around 150–180 ms (STIR—short tau inversion recovery). It is independent from magnetic field inhomogeneities regularly encountered in breast MRI and provides high fluid contrast [7] . The MRI Breast Imaging Reporting and Data System (BI-RADS) lexicon provides defined descriptors for breast lesions in order to standardize reporting of breast pathology [8] . T2 imaging features like edema are included as additional criteria in this lexicon without specification on what imaging technique should be used to evaluate this feature. Diagnostic parameters of edema have been assessed by different groups applying different MRI techniques [5,6,9,10] . As results are varying and as there is no agreement on which T2 imaging technique should be applied for T2w imaging of the breast, the present study was conducted in order to compare the performance of two widely used T2w MRI techniques for the assessment of perifocal edema in breast magnetic resonance imaging. 2 Materials and methods 2.1 Patients and lesions Three hundred twenty-one consecutive patients from a 22 month period were included in this single-center retrospective ethical review board approved investigation. All were referred to MR-mammography in the same department due to conventional BI-RADS III-VI findings. They received definite histopathological diagnosis of imaging detected lesions as a reference standard performed by a board certified breast pathologist (Blinded) at the department of pathology of our university. Specimens were obtained either by open surgery, core biopsy or vacuum biopsy. Further selection criteria were absence of presurgical chemotherapy or biopsy, resulting in 108 malignant and 107 benign lesions (mean patient age 52.4 ± 12.6 (SD), median 51 years). All patients gave their written informed consent to the examination. Note that some lesions in the present investigation overlap with previous articles in different radiological context [11,12] . 2.2 Magnetic resonance imaging All MR-mammographies were performed on a 1.5 T Magnetom Sonata system using the dedicated bilateral phased array breast coil (Siemens, Erlangen, Germany). Patients were positioned head first in prone position. In concordance with international guidelines, the imaging protocol in axial orientation included a dynamic T1 weighted gradient echo sequence (FLASH 2D, parallel imaging GRAPPA factor 2, repetition time (TR) 113 ms, echo time (TE) 5 ms, flip angle 80°, matrix 384 pixels, field of view (FOV) 350 mm, slice thickness 3 mm, time of acquisition 1 min). After one precontrast scan, 0.1 mmol/kg bodyweight of gadopentetate dimeglumine (Gd-DTPA, Magnevist, Bayer HealthCare, Leverkusen, Germany) was administered intravenously as a rapid bolus (3 ml/s) by an automatic injector (Spectris, Medrad, Pittsburgh, USA). This procedure was immediately followed by 20 ml of a saline solution. 30 s after contrast agent injection, postcontrast dynamic imaging was continued under identical tuning conditions until 7 measurements in 1 min intervals were acquired. The imaging protocol further included a T2-weighted Turbo-Spin-Echo sequence (TR 8900 ms, effective TE 207 ms, flip angle 90°, matrix 512 pixels, FOV 350 mm, slice thickness 3 mm) and a Turbo Spin Echo Inversion Recovery sequence with magnitude reconstruction (STIR, TR 8420 ms, effective TE 70 ms, inversion time (TI) 150 ms, flip angle 180°, matrix 256 pixels, FOV 350 mm, slice thickness 3 mm), each in identical slice positions and axial orientation. 2.3 Data analysis All images were read by two radiologists experienced in MR-mammography blinded to all clinical data (WAK and PATB). Reading was performed in consensus applying standardized reading conditions. Precontrast, early and late postcontrast images with their corresponding subtractions, T2w-TSE and STIR images were displayed together at fixed window/center grayvalue settings. Edema was defined as high signal intensity brighter than fat both on T2w-TSE and STIR images which was not caused by cysts. Presence of perifocal edema was rated according to an ordinal scale: 0 = not present, 1 = little, 2 = intermediate, and 3 = distinct. An example for perifocal edema is given in Fig. 1 . Lesion size was assessed using electronic calipers and ordinally classified in 1 = ≤5 mm, 2 = 6–10 mm, 3 = 11–20 mm, 4 = 21–30 mm, 5 = 31–50 mm, and 6 = >50 mm. 2.4 Statistical analysis Ordinal presence of edema was compared with categorial lesion features using Pearson's rank correlation coefficient. Diagnostic parameters calculated together with their corresponding 95% confidence intervals were: sensitivity, specificity, positive and negative predictive value and positive and negative likelihood ratio. Sequence comparison regarding perifocal edema detection was performed using Visual Grading Characteristics (VGC) analysis [13] . Statistical analyses were performed using Medcalc 11.3 (Mariakerke, Belgium) and Meta-DiSc 1.4 ( http://www.hrc.es/investigacion/metadisc.html ) in a Windows XP SP3 environment. 3 Results Median size of both benign and malignant lesions was category 3, 11–20 mm. Presence of edema was significantly correlated with lesion type (T2w-TSE: ρ = 0.525, 95% CI: 0.421–0.616, P < 0.001; STIR: ρ = 0.528, 95% CI: 0.425–0.618, P < 0.001) as it was associated almost exclusively with malignant lesions (for details and resulting diagnostic parameters c.f. Tables 1 and 2 ). Invasive ductal cancer was the malignant subgroup with highest observed frequency of perifocal edema whereas edema surrounding benign lesions accumulated in the inflammatory changes subgroup (c.f. Table 3 ). Regarding lesion characteristics, significant correlations were found between edema and lesion size in malignant lesions (T2w-TSE: ρ = 0.241, 95% CI: 0.054–0.412, P = 0.012; STIR: ρ = 0.219, 95% CI: 0.031–0.393, P = 0.023) as well as grading (T2w-TSE: ρ = 0.230, 95% CI: 0.043–0.401, P = 0.017; STIR: ρ = 0.227, 95% CI: 0.040–0.399, P = 0.018). No correlations were found between lesion size of benign lesions, hormonal receptor status or Her-2/neu receptor status and presence of edema both in T2w-TSE and STIR sequences ( P > 0.05, respectively). VGC analysis revealed a non-significant area under the VGC curve (AUC: 0.509, S.E. 0.026, P = 0.814) and thus no significant differences between both sequences for detecting presence of edema (c.f. Fig. 2 ). 4 Discussion According to our results, both T2w-TSE without fat saturation and STIR sequences can be used the assessment of perifocal edema in MR-mammography. Only minor differences between both sequences were observed. Edema was identified as a highly specific finding of malignancy (positive likelihood ratio T2w-TSE: 9.58 and STIR: 11.29) and, furthermore, showed a positive correlation with tumor grading. Our findings in general are in congruence with previous reports on edema assessment in MRM, describing an association with malignant breast lesions. However, prevalence of this morphologic feature in previous reports differs to some degree. Renz et al. using a similar T2w-TSE technique reported a high prevalence of perifocal edema in locally advanced (71.2%) and inflammatory (66.7%) breast cancers which is in good agreement with our own findings in lesions >2 cm in size, whereas Baltzer et al. reported a sensitivity of 33.5% and a specificity of 93.9% [5,9] . We explain these differences by the lower spatial resolution used in this previous investigation [5] . Comparing our findings with these previous reports shows that all available data suggest an association between malignant lesion size and edema [5,9] . As presence of edema is attributed to proteolytic activity and increased vascular permeability as well as deficient lymphatic vessels in malignant neoplasms, these effects are clearly more prominent in larger lesions [3,5,9,14,15] . However, edema is less frequent but as specific for malignancy in small malignant lesions, thus exhibiting diagnostic value also in early stages of disease. Another report by Telesca et al. on edema assessment using a STIR sequence showed a substantial higher sensitivity but low specificity of edema [6] . Especially perifocal edema was frequently observed in benign lesions. STIR sequences provide added T1 and T2 contrast as the short T2 relaxation time of fat requires short inversion times [7] . Consequently, all other tissues show negative longitudinal magnetization leading to added T1-effects upon T2-weighted STIR imaging. Water has both long T1 and long T2 relaxation times. Consequently, water containing structures like non fibrotic parenchyma show high signal intensity in STIR imaging and may explain the very high frequency of especially diffuse edema in the report by Telesca et al. (also c.f. Fig. 1 ). Although STIR sequences provide robust fat saturation and very good contrast, SNR and, thus, spatial resolution is limited [7] . In this context, pitfalls of perifocal edema assessment have to be mentioned. These include presence of cysts, increased fluid content of breast tissue in young and lactating women and dilated ducts due to ductal obstruction by benign lesions, i.e. papilloma [16] . Such findings may be mistaken for edema and thus decrease specificity of this descriptor [6] . Considering ductal anatomy as well as morphology of high signal intensity areas on T2-weighted imaging is helpful to avoid the described pitfalls in edema assessment. According to our results, STIR imaging was not superior to T2w-TSE imaging for perifocal edema detection. This may be due to increased background tissue signal and a lower spatial resolution in STIR imaging compared to T2w-TSE imaging without fat saturation. Our results suggest that T2 weighted imaging provides valuable diagnostic information for differential diagnosis of breast lesions independent from the sequence used. Some limitations of the present study have to be addressed. The chosen approach of a consensus reading by two experienced radiologists does not allow for investigation of interobserver variability and learning effects. The applied study design compared both T2-weighted sequences in the same reading session. This was done in order to compare image findings intraindividually as the aim was to identify whether edema could be better depicted in either sequence. Visual Grading Characteristics analysis revealed no significant differences in categorical edema assessment. Consequently, no sequence investigated seems to be more sensitive for detection of present edema. Furthermore, our data suggest that there is no value of classifying the magnitude of edema. Distinct edema did not show higher specificity than simple presence of edema (c.f. Table 1 ). However, a quantitative approach with semiautomatic computer assisted detection and classification of edema might provide additional information compared to subjective edema assessment. 5 Conclusion In conclusion, perifocal edema is a very specific (94.4%) feature of malignancy showing moderate sensitivity (53.7%). Although there is a weak correlation with lesion size, this feature is also of high diagnostic value in small lesions due to its high specificity. Both investigated T2 weighted sequences can be used for edema assessment in breast MRI as no substantial differences were observed in VGC analysis. References [1] S.L.J. James D.M. Panicek A.M. Davies Bone marrow oedema associated with benign and malignant bone tumours Eur J Radiol 67 July (1) 2008 11 21 [2] N. Morakkabati C.C. Leutner A. Schmiedel H.H. Schild C.K. Kuhl Breast MR imaging during or soon after radiation therapy Radiology 229 December (3) 2003 893 901 [3] T.L. Adair-Kirk R.M. Senior Fragments of extracellular matrix as mediators of inflammation Int J Biochem Cell Biol 40 6–7 2008 1101 1110 [4] M. Yamaguchi V. Jadhav A. Obenaus A. Colohan J.H. Zhang Matrix metalloproteinase inhibition attenuates brain edema in an in vivo model of surgically-induced brain injury Neurosurgery 61 November (5) 2007 1067 1075 discussion 1075–1076 [5] P.A.T. Baltzer F. Yang M. Dietzel Sensitivity and specificity of unilateral edema on T2w-TSE sequences in MR-mammography considering 974 histologically verified lesions Breast J 16 June (3) 2010 233 239 [6] M. Telesca F. Pediconi V. Dominelli F. Vasselli C. Catalano R. Passariello Does edema differ between benign and malignant breast lesions? Role of T2-weighted STIR imaging Insights Imaging 1 1 2010 S306 S307 [7] G.M. Bydder I.R. Young MR imaging: clinical use of the inversion recovery sequence J Comput Assist Tomogr 9 August (4) 1985 659 675 [8] D.M. Ikeda N.M. Hylton K. Kinkel Development, standardization, and testing of a lexicon for reporting contrast-enhanced breast magnetic resonance imaging studies J Magn Reson Imaging 13 June (6) 2001 889 895 [9] D.M. Renz P.A.T. Baltzer J. Böttcher Inflammatory breast carcinoma in magnetic resonance imaging: a comparison with locally advanced breast cancer Acad Radiol 15 February (2) 2008 209 221 [10] D.M. Renz P.A.T. Baltzer J. Böttcher Magnetic resonance imaging of inflammatory breast carcinoma and acute mastitis. A comparative study Eur Radiol 18 November (11) 2008 2370 2380 [11] Benndorf M, Baltzer PA, Kaiser WA, Assessing the degree of collinearity among the lesion features of the MRI BI-RADS lexicon. Eur J Radiol 2010.[epub ahead of print]. [12] P.A. Baltzer M. Benndorf M. Dietzel M. Gajda I.B. Runnebaum W.A. Kaiser False-positive findings at contrast-enhanced breast MRI: a BI-RADS descriptor study AJR Am J Roentgenol. 194 Jun (6) 2010 658 663 [13] M. Båth L.G. Månsson Visual grading characteristics (VGC) analysis: a non-parametric rank-invariant statistical method for image quality evaluation Br J Radiol 80 March (951) 2007 169 176 [14] J. Beltran D.C. Simon W. Katz L.D. Weis Increased MR signal intensity in skeletal muscle adjacent to malignant tumors: pathologic correlation and clinical relevance Radiology 162 January (1) 1987 251 255 [15] C. Munaut A. Noël O. Hougrand J. Foidart J. Boniver M. Deprez Vascular endothelial growth factor expression correlates with matrix metalloproteinases MT1-MMP, MMP-2 and MMP-9 in human glioblastomas Int J Cancer 106 October (6) 2003 848 855 [16] P.A.T. Baltzer C.G.N. Kaiser M. Dietzel Value of ductal obstruction sign in the differentiation of benign and malignant breast lesions at MR imaging Eur J Radiol 75 August (2) 2010 e18 21