- Email: [email protected]
Fat Necrosis of the Breast
ARTICLE IN PRESS
Original Investigation
Fat Necrosis of the Breast: Magnetic Resonance Imaging Characteristics and Pathologic Correlation Hebatallah Hassan Mamdouh Hassan, PhD, MD, Amr Magdi El Abd, PhD, MD, Amany Abdel Bary, PhD, MD, Nagy N. N. Naguib, PhD, MD Rationale and Objectives: This study aims to describe the magnetic resonance imaging (MRI) features of fat necrosis on magnetic resonance mammography, which may downstage a suspicious lesion to a merely benign finding. Materials and Methods: This prospective study included 82 female patients (mean age 50 years) who were diagnosed to have suspicious lesions by mammography, ultrasonography or both. All patients underwent MRI including diffusion-weighted imaging and spectroscopy. Image postprocessing and analysis included signal intensity, enhancement characteristics, diffusion restriction, and spectroscopic analysis. All patients underwent histopathological analysis for confirmation. Sensitivity, specificity, positive predictive value (PPV), and negative (NPV) predictive value were calculated. Results: To label a lesion as fat necrosis on MRI analysis, presence of fat signal in a lesion revealed sensitivity of 98.04%, specificity of 100%, PPV of 100%, and NPP of 96.88%, whereas nonenhancement of the lesion itself revealed sensitivity of 96.08%, specificity of 100%, PPV of 100%, and NPP of 93.94%. However, adding both the nonrestriction on diffusion analysis and the lack of tCholine at 3.22 ppm increased the sensitivity and specificity to 100%, as well as PPV of 100% for fat necrosis and hence a NPV for malignancy of 100%. Conclusions: MRI proved to be of value in differentiating fat necrosis from malignancy based on the molecular composition of fat necrosis, clearly depicted by MRI without the need for invasive confirmation by biopsy. Key Words: Magnetic resonance imaging of fat necrosis; recurrence; breast cancer; diffusion of fat necrosis; spectroscopy. © 2018 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
F
at necrosis is a benign non–suppurative inflammatory process of adipose tissue (1,2), initially described in the breast in 1920s (1). It is described as “an innocent lesion” in medical literature labeled as BI-RADS 2, which stands for totally benign breast lesion if it met its classical oil cyst form on mammogram. Nonetheless, it gained its notorious reputation and clinicians thrive to diagnose it accurately, as it is the number one differential diagnosis of an early breast Acad Radiol 2018; ■:■■–■■ From the Department of Diagnostic and Interventional Radiology, Faculty of Medicine, Alexandria University, Champollion Street, El Azareeta, Alexandria, Egypt (H.H.M.H., A.M.E.A., N.NN.N.); Department of Medical Pathology, Faculty of Medicine, Alexandria University, Egypt (A.A.B.); Institute for Diagnostic and Interventional Radiology, Johann Wolfgang Goethe University Hospital, Frankfurt am Main, Germany (N.NN.N.). Received September 7, 2017; revised December 25, 2017; accepted December 27, 2017. Institutional review board approval was obtained by the Ethics Committee for Research and Postgraduate Studies of the Faculty of Medicine and University Hospital, Alexandria University. Address correspondence to: H.H.M.H. e-mail: [email protected] © 2018 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.acra.2017.12.019
cancer (3,4). Before the era of imaging and tissue biopsy, the treatment of choice was a wide resection of the breast (1). Even with the advent of the different breast imaging procedures, fat necrosis still displays a wide spectrum of morphologic criteria on the different imaging modalities. This depends on the pathological stage of fat necrosis process, which depends on the balance of fat content and the degree of inflammation and fibrosis of the lesions (5). The classical appearance of fat necrosis would be fat-containing lesion that appears as lucent lesion on mammography, which might be associated with curvilinear or rim calcification termed oil cyst. On ultrasound, it appears either as anechoic or a low-level echogenic cyst within a fat lobule that might be hyperechoic due to edema or isoto hypoechoic due to fibrosis (1,3,5–7). The consensus seems to be that mammograms are probably the best way to investigate fat necrosis because mammography can depict fat-containing lesions by its characteristic lucency better than ultrasound (1,3,5–7). What is of concern is the worrisome end of the spectrum where fibrosis is so florid that it presents as speculated infiltrative 1
Academic Radiology, Vol ■, No ■■, ■■ 2018
HASSAN ET AL
mass lesion or architectural distortion, which again might be accompanied by microcalcification, hence marking it as carcinoma until proved otherwise (1,3,5–7). To differentiate between carcinoma and fat necrosis on imaging basis without the need for a biopsy, magnetic resonance imaging (MRI) of the breast with special emphasis on diffusion and spectroscopic analysis is put to the test. Our aim is to describe the collective MRI features of fat necrosis on magnetic resonance mammography (MRM) and correlate with histopathology to detect specificity and positive predictive value of such a technique in differentiating fat necrosis from carcinoma. MATERIALS AND METHODS The ethics committee of the university hospital and the study review committee of our institute approved this prospective study. A written informed consent was obtained from all patients included in the current study. Patients
Over a course of 1 year starting March 2014 until February 2015, all patients who presented to our institution with suspicious lesions, hence labeled as BI-RADS 4 and 5 lesions, were included, namely, 1. On mammography: as irregular masses with spiculated margin, high-density postoperative scar, developing asymmetry or architectural distortion, as well as coarse heterogeneous calcifications (which might be equivocal for early dystrophic calcifications). 2. On ultrasonography: as irregular lesions with spiculated margins, hypoechoic pattern with special emphasis on heterogeneous or complex cystic and solid ones in addition to posterior shadowing or combined pattern as well as associated architectural distortion and or probe hardness. This was decided through a consensus of two radiologist with 11 and 9 years of experience in mammographic reading. All patients underwent both MRI and biopsy. The following were excluded: 1. Patients with provisional diagnosis of fat necrosis on mammography, mammosonography or both, labeled as BIRADS 2 and 3. This included oil cysts presenting whether as central low density within a scar or oil cyst with rim calcification on mammography, or a cystic lesion with or without low level echoes within a hyper- or isoechoic fat lobule or within the scar itself on sonography. This accounted for 274 patients. 2. Patients with general contraindications for MRI (eg, patients with cardiac pacemaker). 3. Patients with contraindication to MRI contrast agent. 4. Patients with contraindication to biopsy (eg, bleeding tendency and abnormal coagulation time). 2
5. Patients who refused to undergo MRI. 6. Patients who refused biopsy. At the end of the year, we ended up with 82 female patients (mean age ± SD: 50.10 ± 10.55, age ranging between 34 and 76 years), with 58 being suspected for postmanagement recurrence, whereas the rest were newly diagnosed lesions. MRI
All patients underwent MRI on Siemens 1.5 T Magnetom Avanto machine (Erlangen, Germany) using a four-channel breast coil. A histopathological analysis was carried out for confirmation of all cases. MRI Protocols and Technique The imaging protocol consisted of an initial rapid gradientecho scout localization sequence acquired in all three orthogonal planes through both breasts. Noncontrast sequences of the breasts, axilla, and chest wall were acquired in the axial plane notably T2-weighted, fat-saturated sequence with repetition time (TR) (5600 ms), echo time (TE) (59 ms), field of view (FOV) (270–340 mm), matrix (320*314), slice thickness (4 mm), and gap (20% = 0.8 mm) in the axial plane. T1-weighted, non–fat-saturated sequence with TR (8.6 ms), TE (4.7 ms), FOV (270–340), matrix (448*323), slice thickness (1 mm), and gap (−10% = −0.1 mm) in the axial plane with trial to visualize the axillae as we lack a separate axillary coil. Diffusion-weighted echo-planar imaging sequence is applied before contrast administration with TR (4800 ms), TE (98 ms), FOV (270–340), matrix (192*192), slice thickness (4 mm), and gap (50% = 2 mm) in the axial plane, repeated with b-values of 0, 400, and 800, and automatically computer-generated apparent diffusion coefficient (ADC) map. For optimization of the image quality, we ensure automatic presequence shimming using parallel imaging and good fat suppression in addition to previously aforementioned appropriate b-value selection. Three-dimensional T1-weighted gradient-echo sequence was carried out next, with a repetition time of 60 seconds and intravenous administration of 0.2 mmol/kg of gadolinium chelate. Imaging was performed with an FOV of 270– 340 mm over a minimum matrix of 448*322 and slice thickness of 1 mm or less with no gap and an overlap of around 10% using a TR (4.3) and TE (1.3). Spectroscopy of enhancing lesions is performed using single voxel spectroscopy, with the least voxel size being 1 cm3 to have good SNR, which should be >2. Internal water referencing for quantifying total choline (tCholine) was used (8,9). During placement of voxel for spectroscopic analysis, any neighboring cysts, necrotic regions, and radiographic markers or calcific foci were avoided so that the quality of shimming is not degraded (8,9). Image Post Processing Image postprocessing techniques, using Syngo Siemens Medical Solutions software, were applied consisting of three parts: image subtraction, the creation of time-enhancement curves, and lesion
Academic Radiology, Vol ■, No ■■, ■■ 2018
evaluation for the degree of suspicion by generating ADC maps for assessment of lesion’s diffusion and detecting tCholine. During placement of region of interest for diffusion analysis, any neighboring cysts or necrotic regions were avoided to not misread the ADC values generated due to averaging. The used peaks for spectroscopic analysis were 3.18–3.32 ppm for true choline and 4.7 ppm for water (8,9). Lesion Evaluation and Interpretation of Examination Two radiologists with 11 and 9 years’ experience in breast imaging carried out image interpretation in consensus. Meanwhile, they were blinded to the related mammography and sonography images of the patients, reading only the MRI. A systematic approach was adopted to review the MRM studies that consisted of morphologic evaluation, with special emphasis on the lesion’s signal intensity and enhancement analysis. An ADC cut-off value of 1 × 10−3 mm2/s was set for diffusion restriction, where lesions exhibiting values below would be considered restricted, whereas a cut-off value of 1.5 × 10−3 mm2/s was set for nonrestriction and whatever above would be considered nonrestricted, and the interval in between would be considered an overlap zone (10–15). Various research tried to narrow it down to cut-off values of 1.3 (12) and 1.26 (15) × 10−3 mm2/s, in addition to a meta-analysis study suggesting a cut-off value of 1.23 × 10−3 mm2/s (16). For the current study, a cut-off value starting at 1.3 was assigned for nonrestricted diffusion. True choline should be detected at 3.22 ppm, whereas other cellular by-products would be detected at 3.28 and above. Hence, a choline peak at 3.27–3.28 ppm (assigned to glycerophosphocholine, taurine, and myo-inositol) was defined as benign, whereas a peak resonance at 3.22–3.23 ppm (assigned to phosphocholine) was defined as malignant, still the use of a cut-off value is controversial up until now (8,9). Lastly, a BI-RADS class was assigned to the study at hand before the histopathological correlation and it was compared to the provisional BI-RADS class chosen on the conventional mammography or sonography. Histopathological Analysis
All lesions were biopsied under sonographic guidance using semiautomatic core biopsy needles (Geotek, Geotek Medikal Ltd., Ankara, Turkey) of 16 and 14 gauge according to lesion’s size and were processed and reviewed by an experienced pathologist (7 years). Data and Statistical Analysis
Statistical analysis of the generated data from the individual MRM sequences will entail sensitivity, specificity, and both PPV and NPV both individually and compared to each other to scrutinize the feasibility and practicality of each sequence individually and in adjunct to others using SPSS software version 20. (IBM SPSS Statistics for Windows, Version 20.0. IBM Corp. 2011, Armonk, N.Y., USA)
MRI CHARACTERISTICS OF FAT NECROSI BREAST
RESULTS Over the course of 1 year, 82 patients with mammosonographic suspicious lesions underwent MRI and histopathological correlation. Fat necrosis was diagnosed in 51 cases, where 49 of 51 cases were postoperative in nature, whereas malignant growth was detected in 31 cases, with 9 of 31 cases being recurrence on the postoperative scar. MRI Findings
Morphology and Signal Intensity Although both pathologically proven that fat necrosis and cancerous lesions had a spiculated outline, still the difference was in the signal intensity of the lesion itself. Central T1 hyperintensity of fat was detected in 50 of 51 cases of fat necrosis; this was proven by marked nulling after fat saturation on both T2- and T1-fat suppression sequences. Note that the suppression of the lesions was even more than the normal breast fat, whereas the remaining patient showed markedly hypointense focal lesion up to signal void on both T1 non–fat-saturated and T2 fat-suppressed sequences. The remaining 31 of 82 patients with malignant lesions revealed no fat signal. Associated surrounding fibrosis of variable degree and edematous changes were observed in 76 of 82 cases. Enhancement Characteristics Enhancement pattern was divided into nonenhancing and enhancing lesions, where again enhancing lesions were subdivided into rim enhancement, homogeneous-, or heterogeneousenhancing lesions as per the classification of the Bi-RADS lexicon. Regarding the 51 pathologically proven fat necrosis lesions, the center of the lesion (fat) did not show enhancement in 49 of 51 patients. Nonenhancement of the surrounding fibrosis was encountered in 11 of 51 patients, whereas rim enhancement of the surrounding fibrosis was encountered in 38 of 51 patients. A heterogeneous enhancement was identified in the remaining two patients, which was of type I progressive enhancement. All patients with cancerous lesions (31 of 81) revealed enhancing lesions regardless of the curve pattern, homogeneous enhancement in 19 of 31 patients, and heterogeneous in 11 of 31 patients, all exhibited type III rapid wash in washout pattern apart from the three lesions of type I pattern. Diffusion-weighted Imaging (DWI) and Spectroscopic Analysis Cases of fat necrosis (51 of 82) exhibited no diffusion restriction, with ADC values ranging from 1.3 to 2.4 × 10 −3 mm 2 /s, with a mean value ± SD of 1.63 ± 0.26 × 10−3 mm2/s. The remaining 31 cases revealed lesions with ADC values < 1 × 10−3 mm2/s (mean value ± SD 0.84 ± 0.22 × 10−3 mm2/s) reflecting frank restriction. Spectroscopic analysis was performed only in cases exhibiting enhancement with enough sample size to fit the voxel placement, hence 46 of 82 cases were scanned. As for tCholine trace at 3.22 ppm (true phosphocholine), only 27 of 46 cases showed positive trace, 8 of 46 cases revealed positive tCholine 3
Academic Radiology, Vol ■, No ■■, ■■ 2018
HASSAN ET AL
TABLE 1. MRI Findings of Pathologically Proven Cases of Fat Necrosis MRI Characteristics T1 intensity T2 fat-suppressed intensity Lesion's enhancement
Diffusion-weighted imaging
Spectroscopic analysis
Hyperintense Hypointense (signal void) Black hole effect (markedly hypointense) Signal void Enhancing rim with nonenhancing center Nonenhancement of both center and surrounding fibrosis Diffuse enhancement Nonrestricted (≥1.5 × 10−3 mm2/s) Overlap zone (>1.3 × 10−3 mm2/s) (<1.3 × 10−3 mm2/s) Restricted (≤1.0 × 10−3 mm2/s) Positive (at 3.22 ppm) Positive (>3.28 ppm) Negative (at 3.2 ppm and >3.28 ppm)
Number of Cases
Total
50 1 50 1 38 11 2 42 9 0 0 0 4 11
51 51 51
51
15
MRI, magnetic resonance imaging.
trace > 3.28 ppm (other by-products than phosphocholine), whereas the remaining 11 of 46 cases showed negative tCholine trace on both peaks. BI-RADS Classification Reading all parameters together, 58.54% (48 of 82) of lesions were assigned BI-RADS 2 class, 3.66% (3 of 82) were assigned BI-RADS 3, which was in concordance with the histopathological results of fat necrosis, whereas the remaining 37.80% (31 of 82) were assigned BI-RADS 4 and 5, which matched the histopathological results of cancerous lesions. It has to be stressed on that diffusion and spectroscopic analysis are not established parameters of the BI-RADS system yet, and lesions were assigned according to the morphology and kinetic analysis into the different BI-RADS classes; still, diffusion and spectroscopic analysis endorsed the final selection of the BI-RADS class without the need to overclassify lesions. Radiologic-Pathologic Correlation of Cases with Fat Necrosis
Morphology and Signal Intensity A fat signal was reproducible in 98% (50 of 51) of the lesions (Table 1, Fig 1), the remaining case lacking fat signal was replaced by signal voids to be pathologically proven hemorrhage on a background of florid fibrosis (Fig 2). Enhancement Characteristics An enhancement was encountered only in 2 of 51 cases (4%), which was heterogeneous, still exhibiting type I curve of progressive enhancement (Table 1). Histopathologically, this is explained by the coexisting different stages of inflammatory changes and fibrosis (Fig 2). Nearly three quarter or 74.5% (38 of 51) of the cases revealed enhancement of the surrounding fibrosis, again of type I, none showed type II or III curve patterns, whereas the remaining 21.5% (11 of 51) revealed no enhancement of the surrounding fibrous reac4
tion, pathologically proven to be mature fibrous tissue, including the lesion lacking fat signal (Fig 1). DWI and Spectroscopic Analysis None of the lesions of fat necrosis revealed diffusion restriction, even lesions showing heterogeneous enhancement had ADC values of 1.6 and 1.9 × 10−3 mm2/s, respectively (Table 1, Figs 1, 2). Spectroscopic analysis was performed to 15 of 51 cases of fat necrosis; 2 heterogeneously enhancing lesions and 13 cases with extended surrounding enhancing fibrosis. None revealed choline trace at 3.22 ppm, 4 of 15 (26.67%) displayed positive tCholine trace > 3.28 ppm, including one of the heterogeneously enhancing lesions at 3.34 ppm, whereas the remaining 11 of 15 (73.33%) cases showed negative tCholine trace, including the second heterogeneously enhancing lesion (Fig 2). Diagnostic Performance
To label a lesion as fat necrosis on MRI analysis, detection of fat signal in a lesion revealed sensitivity of 98.04%, specificity of 100%, positive predictive value (PPV) of 100%, and negative predictive value (NPP) of 96.88%. Nonenhancement of the lesion itself revealed sensitivity of 96.08%, specificity of 100%, PPV of 100%, and NPV of 93.94%. However, adding the nonrestriction on diffusion analysis increased the sensitivity and specificity to 100% and PPV of 100% for fat necrosis and hence an NPV for malignancy of 100%. When putting spectroscopic analysis to the test, the lack of tCholine at 3.22 ppm yielded reproducible results as in diffusion analysis, namely, sensitivity, specificity, and PPV of 100% for fat necrosis, and in return a NPV of 100% for malignancy. DISCUSSION The appearance of fat necrosis is variable according to the fat content and the degree of inflammation and fibrosis of the lesions (1–5). MRI is considered as a problem-solving tool as it depends on detection of the molecular composition of the lesion,
Academic Radiology, Vol ■, No ■■, ■■ 2018
MRI CHARACTERISTICS OF FAT NECROSI BREAST
Figure 1. A 44-year-old female patient with a history of right breast cancer managed surgically followed by chemo- and radiotherapy on periodic follow-up. Ultrasound examination (a) identified a linear hypoechogenicity matching with postoperative scar; 2 years later, follow-up ultrasound (b) revealed irregular stellate hypoechoic lesion, raising the suspicion of malignancy. Axial T1 non–fat-suppressed (c) and T2 fat-suppressed (d) images revealed marked focal fibrosis along the operative scar associated with few internal fat foci reflecting end-stage fat necrosis with maturing fibrous tissue. Postcontrast T1 subtracted (e) and MIP (f) images show negligible enhancement of the fibrosis with central nonenhancing fat foci confirming fibrosis maturation. Diffusion-weighted image (g) with b-value of 800 showed no evidence of diffusion restriction confirming the nonrestricted values on ADC map (h), endorsing the BI-RADS 2 class assigned to the lesion based on morphology and kinetics analysis and excluding suspicion of recurrence.
facilitating the depiction of fat, which is still the most reliable imaging sign for the diagnosis of fat necrosis (2,7). Still, the current literature (2,7) solely describes the various morphologic changes of fat necrosis. Unfortunately, there will be instances in which fat necrosis is indistinguishable from cancer on MRI. Lesions with no gross lipid and a preponderance of fibrosis can appear spiculated and nearly solid, suggesting cancer rather than fat necrosis (2,7).
Hence, we came up with the idea of putting MRI of the breast to the test by reviewing not only the morphologic but also the new sequences that depend on the biological behavior of the lesion such as DWI and spectroscopic analysis in an attempt to use it to differentiate between breast cancer and fat necrosis. All lesions in the current study were nonclassical for fat necrosis on both mammography and sonography, this might be 5
HASSAN ET AL
Academic Radiology, Vol ■, No ■■, ■■ 2018
Figure 2. A 54-year-old female patient with a history of right breast cancer managed surgically followed by chemo- and radiotherapy on follow-up. Mammographic MLO view of the right breast (a) revealed upper outer quadrant focal asymmetry, and ultrasound (b) identified a suspicious focus within the operative bed. Axial T1 non–fatsuppressed (c) and T2 fat-suppressed (d) images revealed a markedly hypointense focus on both sequences along the operative scar associated with surrounding edema. No fat signal is identified, still marked signal void of hemosiderin deposition is the most likely cause. Postcontrast subtracted image for mean curve analysis (e) revealed focal enhancement of type I b progressive enhancement with delayed washout. No central nonenhancement identified, making the diagnosis equivocal for recurrence; still no choline trace was identified on spectroscopic analysis (f), with the value of tCholine being zero at the 3.22 ppm point along the scale, endorsing the BIRADS 3 class assigned to the lesion based on morphology and kinetics analysis and excluding the suspicion of recurrence. Diffusion-weighted image (g) with b-value of 800 and ADC map (h) showed no evidence of diffusion restriction, confirming the spectroscopic analysis and excluding recurrence. Low power microscopic examination of the lesion (i) shows multiple communicating fat cysts, hemorrhage, and a diffuse mixed inflammatory infiltrate (hematoxylin and eosin [H&E] ×100), higher power examination (j) shows large fat cyst surrounded by necrosis entangling karyorrhectic debris and a mixed inflammatory infiltrate (H&E × 200). High power examination (k) confirms the inflammatory infiltrate formed by large foamy histiocytes (arrows), neutrophils, lymphocytes, and hemosiderinladen macrophages (H&E × 400).
attributed to the fact that the fibrosis was taking the upper hand is so florid, even experienced radiologists could not confidently diagnose it as fat necrosis especially in the setting of the postmanaged breast. Of 82 examined patients, 51 were pathologically proven to be fat necrosis. This was concordant with the final MRIBI-RADS category based on the signal intensity, enhancement characteristic, DWI, and spectroscopic analysis. Regarding the signal intensity of the lesions, 98% (50 of 51) of the pathologically proven fat necrosis lesions detected in this series did show fat signal, which concurs with current literature (1–7,17,18), sometimes termed black hole as they become markedly hypointense compared to the surrounding fat (18). 6
The only case not conforming to this finding, which can be labeled lipid poor, is explained in the literature (1,3–7) by the fact that lack of gross lipid and preponderance of fibrosis within a lesion gives it the appearance of nearly solid, suggesting cancer rather than fat necrosis. Moreover, hemorrhage, which is a described component of the inflammatory changes taking place in fat necrosis, can lead to signal voids on MRI due to its iron content (3,4), both phenomena were proven present during the histopathological analysis of this case. The current literature (1,3–7) finds kinetic analysis to be of little help, as fat necrosis exhibits the full spectrum of benign and malignant enhancement patterns as a result of a balance between the degree of inflammation and maturing fibrosis. This was only observed in 4% (2 of 51) of the cases. In contrast,
Academic Radiology, Vol ■, No ■■, ■■ 2018
MRI CHARACTERISTICS OF FAT NECROSI BREAST
Figure 2. (continued).
96% (49 of 51) of the cases did show nonenhancement of the core of the lesion containing gross fat foci, whereas the surrounding fibrosis enhancement varied according to degree of maturity; this yielded a PPV of 100% for a lesion to be fat necrosis, which is considered as an assuring point (1,3–7,17). The existing studies about fat necrosis appearance by MRI did not address diffusion restriction as sole research. However, DWI has been examined in the context of differentiating benign from malignant breast lesions; a number of these studies (8,10,11,15,19) did show that fat necrosis does not express diffusion restriction. This concurred with the current findings, where none of the lesions showed diffusion restriction. The spectroscopic analysis dedicated to the recognition of fat necrosis is sparse in the literature (8,19); still, the protocol of this study speculated that none of the fat necrosis cases would exhibit choline trace at 3.22 ppm, which concurred with the results. Out of the examined 15 cases, only 26.67%
(4 of 15) displayed tCholine trace > 3.28 ppm explainable as by-products of inflammatory changes, whereas 73.33% (11 of 15) revealed negative tCholine trace at both 3.22 and >3.28 ppm. The remaining 36 cases of fat necrosis were not subjected to spectroscopic analysis due to inadequate sample size or lack of enhancement of the lesion itself. Deducted from the current study, to reach a definitive diagnosis, an algorithm for review of the sequences at hand should be taken into consideration to try and reach out the maximum benefit of the examination. The typical appearance should contain fat signal, which usually exhibits “black hole effect.” Hemorrhagic component appearing as signal voids can be present. Enhancement pattern, if present, is typically rim enhancement of the surrounding fibrosis regardless of the type of curve pattern, although type I would be a plus as fibrosis does enhance late. Sometimes the enhancing rim would be thick due to florid fibrosis giving appearance of heterogeneous 7
HASSAN ET AL
enhancement; the same situation is faced when dealing with active inflammatory process. Hence, reverting to the other sequences, no diffusion restriction and negative at 3.22 ppm or low level of tCholine trace > 3.28 ppm would be correspondingly in favor of fat necrosis and inflammatory changes. The limitation of our study was that most cases positive for fat necrosis were already managed cases; therefore, the probability of postmanagement sequel is undeniably higher than the rate of recurrence. Conversely, this point could be used as an advantage to enroll MRI of the breast as part of the routine examination of challenging managed breast. This is due to that fact that we are dealing with a probable recurrence rate that might be as high as 2.5% per year (17), which is problematic to managed cases of breast cancer, particularly those refusing to undergo biopsy or have a contraindication to biopsy as it will yield desirable results with less invasive maneuver. A second limitation is that there is still no standardized cut-off value along the gray zone of ADC values of 1–1.5 × 10−3 mm2/s to label a lesion as benign or malignant. However, based on several research articles (15,16), the suggested cut-off values based on ROC analysis yielding up to 90% sensitivity and 80% specificity ranged from 1.23 to 1.26, hence we decided to use a slightly higher value of 1.3. Needless to say, one cannot depend solely on DWI as a differentiating point for lesion’s nature and it has to be still read as just a part of the whole study and not as a stand-alone sequence. Additional limitation includes the fact that spectroscopic analysis was done only to cases with enhancing lesions, although this is the standard of care, it consequently resulted in a small sample size of fat necrosis lesions tested. Hence, we propose further studies to be carried out even in lack of enhancement to detect any signal in the surrounding inflammatory reaction to test if end-stage fat necrosis—fibrosis—does in fact lack definite by-products. In conclusion, MRI proved to be of value in differentiating fat necrosis from malignancy based on the molecular composition of fat necrosis, which is clearly depicted by MRI. Moreover, multisequence approach proved to be of value in cases where the classical fat intensity signal could not be detected. Hence, in cases showing the previously described MRI signs of fat necrosis, the need for invasive confirmation by biopsy might not be required.
8
Academic Radiology, Vol ■, No ■■, ■■ 2018
REFERENCES 1. Tan PH, Lai LM, Carrington EV, et al. Fat necrosis of the breast—a review. The Breast 2006; 15:313–318. 2. Chan LP, Gee R, Keogh C, et al. Imaging features of fat necrosis. AJR Am J Roentgenol 2003; 181:955–959. 3. Daly CP, Jaeger B, Sill DS. Variable appearances of fat necrosis on breast MRI. AJR Am J Roentgenol 2008; 191:1374–1380. 4. Taboada JL, Stephens TW, Krishnamurthy S, et al. The many faces of fat necrosis in the breast. AJR Am J Roentgenol 2009; 192:815–825. 5. Kerridge WD, Kryvenko ON, Thompson A, et al. Fat necrosis of the breast: a pictorial review of the mammographic, ultrasound, CT, and MRI findings with histopathologic correlation. Radiol Res Pract. 2015; 2015:613139. doi:10.1155/2015/613139. 6. Chala LF, de Barros N, Moraes P, et al. Fat necrosis of the breast: mammographic, sonographic, computed tomography, and magnetic resonance imaging findings. Curr Probl Diagn Radiol 2004; 33:106–126. 7. Sergi Ganau S, Tortajada L, Escribano F, et al. The great mimicker: fat necrosis of the breast—magnetic resonance mammography approach. Curr Probl Diagn Radiol 2009; 38:189–197. 8. Stanwell P, Gluch L, Clark D, et al. Specificity of choline metabolites for in vivo diagnosis of breast cancer using 1H MRS at 1.5 T. Eur Radiol 2005; 15:1037–1043. 9. Sardanelli F, Fausto A, DiLeo G, et al. In vivo proton MR spectroscopy of the breast using the total choline peak integral as a marker of malignancy. AJR Am J Roentgenol 2009; 192:1608–1617. 10. Partridge SC, DeMartini WB, Kurland BF, et al. Quantitative diffusionweighted imaging as an adjunct to conventional breast MRI for improved positive predictive value. AJR Am J Roentgenol 2009; 193:1716–1722. 11. Kul S, Cansu A, Alhan E, et al. Contribution of diffusion weighted imaging to dynamic contrast-enhanced MRI in the characterization of breast tumors. AJR Am J Roentgenol 2011; 196:210–217. 12. Gonzales MA, Castañon AI, Baltar N, et al. Diffusion weighted MR imaging assessment after breast conservative treatment. Poster session in ECR 2012 Educational Exhibit. March 1–5, 2012; Vienna, Austria. doi:10.1594/ecr2012/C-1065. 13. Şahin C, Arıbal E. The role of apparent diffusion coefficient values in the differential diagnosis of breast lesions in diffusion-weighted MRI. Diagn Interv Radiol 2013; 19:457–462. 14. Hassan HHM, Zahran MHM, Hassan HE, et al. Diffusion magnetic resonance imaging of breast lesions: initial experience at Alexandria University. Alexandria J Med 2013; 49:2652–2672. 15. Hetta W. Role of diffusion weighted images combined with breast MRI in improving the detection and differentiation of breast lesions. Egypt J Radio Nuc Med 2015; 46:259–270. 16. Tsushima Y, Takahashi-Taketomi A, Endo K. Magnetic resonance (MR) differential diagnosis of breast tumors using apparent diffusion coefficient (ADC) on 1.5-T. J Magn Reson Imaging 2009; 30:249–255. 17. Chansakul T, Lai KC, Slanetz PJ. The post-conservation breast: part 2, imaging findings of tumor recurrence and other long-term sequelae. AJR Am J Roentgenol 2012; 98:331–333. 18. Trimboli RM, Carbonaro LA, Cartia F, et al. MRI of fat necrosis of the breast: the “black hole” sign at short tau inversion recovery. Eur J Radiol 2012; 81:e573–e579. 19. Sardanelli F, Carbonaro LA, Montemezzi S, et al. Clinical breast MR using MRS or DWI: who is the winner? Front Oncol 2016; 6:e217.
Original Investigation
Fat Necrosis of the Breast: Magnetic Resonance Imaging Characteristics and Pathologic Correlation Hebatallah Hassan Mamdouh Hassan, PhD, MD, Amr Magdi El Abd, PhD, MD, Amany Abdel Bary, PhD, MD, Nagy N. N. Naguib, PhD, MD Rationale and Objectives: This study aims to describe the magnetic resonance imaging (MRI) features of fat necrosis on magnetic resonance mammography, which may downstage a suspicious lesion to a merely benign finding. Materials and Methods: This prospective study included 82 female patients (mean age 50 years) who were diagnosed to have suspicious lesions by mammography, ultrasonography or both. All patients underwent MRI including diffusion-weighted imaging and spectroscopy. Image postprocessing and analysis included signal intensity, enhancement characteristics, diffusion restriction, and spectroscopic analysis. All patients underwent histopathological analysis for confirmation. Sensitivity, specificity, positive predictive value (PPV), and negative (NPV) predictive value were calculated. Results: To label a lesion as fat necrosis on MRI analysis, presence of fat signal in a lesion revealed sensitivity of 98.04%, specificity of 100%, PPV of 100%, and NPP of 96.88%, whereas nonenhancement of the lesion itself revealed sensitivity of 96.08%, specificity of 100%, PPV of 100%, and NPP of 93.94%. However, adding both the nonrestriction on diffusion analysis and the lack of tCholine at 3.22 ppm increased the sensitivity and specificity to 100%, as well as PPV of 100% for fat necrosis and hence a NPV for malignancy of 100%. Conclusions: MRI proved to be of value in differentiating fat necrosis from malignancy based on the molecular composition of fat necrosis, clearly depicted by MRI without the need for invasive confirmation by biopsy. Key Words: Magnetic resonance imaging of fat necrosis; recurrence; breast cancer; diffusion of fat necrosis; spectroscopy. © 2018 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
F
at necrosis is a benign non–suppurative inflammatory process of adipose tissue (1,2), initially described in the breast in 1920s (1). It is described as “an innocent lesion” in medical literature labeled as BI-RADS 2, which stands for totally benign breast lesion if it met its classical oil cyst form on mammogram. Nonetheless, it gained its notorious reputation and clinicians thrive to diagnose it accurately, as it is the number one differential diagnosis of an early breast Acad Radiol 2018; ■:■■–■■ From the Department of Diagnostic and Interventional Radiology, Faculty of Medicine, Alexandria University, Champollion Street, El Azareeta, Alexandria, Egypt (H.H.M.H., A.M.E.A., N.NN.N.); Department of Medical Pathology, Faculty of Medicine, Alexandria University, Egypt (A.A.B.); Institute for Diagnostic and Interventional Radiology, Johann Wolfgang Goethe University Hospital, Frankfurt am Main, Germany (N.NN.N.). Received September 7, 2017; revised December 25, 2017; accepted December 27, 2017. Institutional review board approval was obtained by the Ethics Committee for Research and Postgraduate Studies of the Faculty of Medicine and University Hospital, Alexandria University. Address correspondence to: H.H.M.H. e-mail: [email protected] © 2018 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.acra.2017.12.019
cancer (3,4). Before the era of imaging and tissue biopsy, the treatment of choice was a wide resection of the breast (1). Even with the advent of the different breast imaging procedures, fat necrosis still displays a wide spectrum of morphologic criteria on the different imaging modalities. This depends on the pathological stage of fat necrosis process, which depends on the balance of fat content and the degree of inflammation and fibrosis of the lesions (5). The classical appearance of fat necrosis would be fat-containing lesion that appears as lucent lesion on mammography, which might be associated with curvilinear or rim calcification termed oil cyst. On ultrasound, it appears either as anechoic or a low-level echogenic cyst within a fat lobule that might be hyperechoic due to edema or isoto hypoechoic due to fibrosis (1,3,5–7). The consensus seems to be that mammograms are probably the best way to investigate fat necrosis because mammography can depict fat-containing lesions by its characteristic lucency better than ultrasound (1,3,5–7). What is of concern is the worrisome end of the spectrum where fibrosis is so florid that it presents as speculated infiltrative 1
Academic Radiology, Vol ■, No ■■, ■■ 2018
HASSAN ET AL
mass lesion or architectural distortion, which again might be accompanied by microcalcification, hence marking it as carcinoma until proved otherwise (1,3,5–7). To differentiate between carcinoma and fat necrosis on imaging basis without the need for a biopsy, magnetic resonance imaging (MRI) of the breast with special emphasis on diffusion and spectroscopic analysis is put to the test. Our aim is to describe the collective MRI features of fat necrosis on magnetic resonance mammography (MRM) and correlate with histopathology to detect specificity and positive predictive value of such a technique in differentiating fat necrosis from carcinoma. MATERIALS AND METHODS The ethics committee of the university hospital and the study review committee of our institute approved this prospective study. A written informed consent was obtained from all patients included in the current study. Patients
Over a course of 1 year starting March 2014 until February 2015, all patients who presented to our institution with suspicious lesions, hence labeled as BI-RADS 4 and 5 lesions, were included, namely, 1. On mammography: as irregular masses with spiculated margin, high-density postoperative scar, developing asymmetry or architectural distortion, as well as coarse heterogeneous calcifications (which might be equivocal for early dystrophic calcifications). 2. On ultrasonography: as irregular lesions with spiculated margins, hypoechoic pattern with special emphasis on heterogeneous or complex cystic and solid ones in addition to posterior shadowing or combined pattern as well as associated architectural distortion and or probe hardness. This was decided through a consensus of two radiologist with 11 and 9 years of experience in mammographic reading. All patients underwent both MRI and biopsy. The following were excluded: 1. Patients with provisional diagnosis of fat necrosis on mammography, mammosonography or both, labeled as BIRADS 2 and 3. This included oil cysts presenting whether as central low density within a scar or oil cyst with rim calcification on mammography, or a cystic lesion with or without low level echoes within a hyper- or isoechoic fat lobule or within the scar itself on sonography. This accounted for 274 patients. 2. Patients with general contraindications for MRI (eg, patients with cardiac pacemaker). 3. Patients with contraindication to MRI contrast agent. 4. Patients with contraindication to biopsy (eg, bleeding tendency and abnormal coagulation time). 2
5. Patients who refused to undergo MRI. 6. Patients who refused biopsy. At the end of the year, we ended up with 82 female patients (mean age ± SD: 50.10 ± 10.55, age ranging between 34 and 76 years), with 58 being suspected for postmanagement recurrence, whereas the rest were newly diagnosed lesions. MRI
All patients underwent MRI on Siemens 1.5 T Magnetom Avanto machine (Erlangen, Germany) using a four-channel breast coil. A histopathological analysis was carried out for confirmation of all cases. MRI Protocols and Technique The imaging protocol consisted of an initial rapid gradientecho scout localization sequence acquired in all three orthogonal planes through both breasts. Noncontrast sequences of the breasts, axilla, and chest wall were acquired in the axial plane notably T2-weighted, fat-saturated sequence with repetition time (TR) (5600 ms), echo time (TE) (59 ms), field of view (FOV) (270–340 mm), matrix (320*314), slice thickness (4 mm), and gap (20% = 0.8 mm) in the axial plane. T1-weighted, non–fat-saturated sequence with TR (8.6 ms), TE (4.7 ms), FOV (270–340), matrix (448*323), slice thickness (1 mm), and gap (−10% = −0.1 mm) in the axial plane with trial to visualize the axillae as we lack a separate axillary coil. Diffusion-weighted echo-planar imaging sequence is applied before contrast administration with TR (4800 ms), TE (98 ms), FOV (270–340), matrix (192*192), slice thickness (4 mm), and gap (50% = 2 mm) in the axial plane, repeated with b-values of 0, 400, and 800, and automatically computer-generated apparent diffusion coefficient (ADC) map. For optimization of the image quality, we ensure automatic presequence shimming using parallel imaging and good fat suppression in addition to previously aforementioned appropriate b-value selection. Three-dimensional T1-weighted gradient-echo sequence was carried out next, with a repetition time of 60 seconds and intravenous administration of 0.2 mmol/kg of gadolinium chelate. Imaging was performed with an FOV of 270– 340 mm over a minimum matrix of 448*322 and slice thickness of 1 mm or less with no gap and an overlap of around 10% using a TR (4.3) and TE (1.3). Spectroscopy of enhancing lesions is performed using single voxel spectroscopy, with the least voxel size being 1 cm3 to have good SNR, which should be >2. Internal water referencing for quantifying total choline (tCholine) was used (8,9). During placement of voxel for spectroscopic analysis, any neighboring cysts, necrotic regions, and radiographic markers or calcific foci were avoided so that the quality of shimming is not degraded (8,9). Image Post Processing Image postprocessing techniques, using Syngo Siemens Medical Solutions software, were applied consisting of three parts: image subtraction, the creation of time-enhancement curves, and lesion
Academic Radiology, Vol ■, No ■■, ■■ 2018
evaluation for the degree of suspicion by generating ADC maps for assessment of lesion’s diffusion and detecting tCholine. During placement of region of interest for diffusion analysis, any neighboring cysts or necrotic regions were avoided to not misread the ADC values generated due to averaging. The used peaks for spectroscopic analysis were 3.18–3.32 ppm for true choline and 4.7 ppm for water (8,9). Lesion Evaluation and Interpretation of Examination Two radiologists with 11 and 9 years’ experience in breast imaging carried out image interpretation in consensus. Meanwhile, they were blinded to the related mammography and sonography images of the patients, reading only the MRI. A systematic approach was adopted to review the MRM studies that consisted of morphologic evaluation, with special emphasis on the lesion’s signal intensity and enhancement analysis. An ADC cut-off value of 1 × 10−3 mm2/s was set for diffusion restriction, where lesions exhibiting values below would be considered restricted, whereas a cut-off value of 1.5 × 10−3 mm2/s was set for nonrestriction and whatever above would be considered nonrestricted, and the interval in between would be considered an overlap zone (10–15). Various research tried to narrow it down to cut-off values of 1.3 (12) and 1.26 (15) × 10−3 mm2/s, in addition to a meta-analysis study suggesting a cut-off value of 1.23 × 10−3 mm2/s (16). For the current study, a cut-off value starting at 1.3 was assigned for nonrestricted diffusion. True choline should be detected at 3.22 ppm, whereas other cellular by-products would be detected at 3.28 and above. Hence, a choline peak at 3.27–3.28 ppm (assigned to glycerophosphocholine, taurine, and myo-inositol) was defined as benign, whereas a peak resonance at 3.22–3.23 ppm (assigned to phosphocholine) was defined as malignant, still the use of a cut-off value is controversial up until now (8,9). Lastly, a BI-RADS class was assigned to the study at hand before the histopathological correlation and it was compared to the provisional BI-RADS class chosen on the conventional mammography or sonography. Histopathological Analysis
All lesions were biopsied under sonographic guidance using semiautomatic core biopsy needles (Geotek, Geotek Medikal Ltd., Ankara, Turkey) of 16 and 14 gauge according to lesion’s size and were processed and reviewed by an experienced pathologist (7 years). Data and Statistical Analysis
Statistical analysis of the generated data from the individual MRM sequences will entail sensitivity, specificity, and both PPV and NPV both individually and compared to each other to scrutinize the feasibility and practicality of each sequence individually and in adjunct to others using SPSS software version 20. (IBM SPSS Statistics for Windows, Version 20.0. IBM Corp. 2011, Armonk, N.Y., USA)
MRI CHARACTERISTICS OF FAT NECROSI BREAST
RESULTS Over the course of 1 year, 82 patients with mammosonographic suspicious lesions underwent MRI and histopathological correlation. Fat necrosis was diagnosed in 51 cases, where 49 of 51 cases were postoperative in nature, whereas malignant growth was detected in 31 cases, with 9 of 31 cases being recurrence on the postoperative scar. MRI Findings
Morphology and Signal Intensity Although both pathologically proven that fat necrosis and cancerous lesions had a spiculated outline, still the difference was in the signal intensity of the lesion itself. Central T1 hyperintensity of fat was detected in 50 of 51 cases of fat necrosis; this was proven by marked nulling after fat saturation on both T2- and T1-fat suppression sequences. Note that the suppression of the lesions was even more than the normal breast fat, whereas the remaining patient showed markedly hypointense focal lesion up to signal void on both T1 non–fat-saturated and T2 fat-suppressed sequences. The remaining 31 of 82 patients with malignant lesions revealed no fat signal. Associated surrounding fibrosis of variable degree and edematous changes were observed in 76 of 82 cases. Enhancement Characteristics Enhancement pattern was divided into nonenhancing and enhancing lesions, where again enhancing lesions were subdivided into rim enhancement, homogeneous-, or heterogeneousenhancing lesions as per the classification of the Bi-RADS lexicon. Regarding the 51 pathologically proven fat necrosis lesions, the center of the lesion (fat) did not show enhancement in 49 of 51 patients. Nonenhancement of the surrounding fibrosis was encountered in 11 of 51 patients, whereas rim enhancement of the surrounding fibrosis was encountered in 38 of 51 patients. A heterogeneous enhancement was identified in the remaining two patients, which was of type I progressive enhancement. All patients with cancerous lesions (31 of 81) revealed enhancing lesions regardless of the curve pattern, homogeneous enhancement in 19 of 31 patients, and heterogeneous in 11 of 31 patients, all exhibited type III rapid wash in washout pattern apart from the three lesions of type I pattern. Diffusion-weighted Imaging (DWI) and Spectroscopic Analysis Cases of fat necrosis (51 of 82) exhibited no diffusion restriction, with ADC values ranging from 1.3 to 2.4 × 10 −3 mm 2 /s, with a mean value ± SD of 1.63 ± 0.26 × 10−3 mm2/s. The remaining 31 cases revealed lesions with ADC values < 1 × 10−3 mm2/s (mean value ± SD 0.84 ± 0.22 × 10−3 mm2/s) reflecting frank restriction. Spectroscopic analysis was performed only in cases exhibiting enhancement with enough sample size to fit the voxel placement, hence 46 of 82 cases were scanned. As for tCholine trace at 3.22 ppm (true phosphocholine), only 27 of 46 cases showed positive trace, 8 of 46 cases revealed positive tCholine 3
Academic Radiology, Vol ■, No ■■, ■■ 2018
HASSAN ET AL
TABLE 1. MRI Findings of Pathologically Proven Cases of Fat Necrosis MRI Characteristics T1 intensity T2 fat-suppressed intensity Lesion's enhancement
Diffusion-weighted imaging
Spectroscopic analysis
Hyperintense Hypointense (signal void) Black hole effect (markedly hypointense) Signal void Enhancing rim with nonenhancing center Nonenhancement of both center and surrounding fibrosis Diffuse enhancement Nonrestricted (≥1.5 × 10−3 mm2/s) Overlap zone (>1.3 × 10−3 mm2/s) (<1.3 × 10−3 mm2/s) Restricted (≤1.0 × 10−3 mm2/s) Positive (at 3.22 ppm) Positive (>3.28 ppm) Negative (at 3.2 ppm and >3.28 ppm)
Number of Cases
Total
50 1 50 1 38 11 2 42 9 0 0 0 4 11
51 51 51
51
15
MRI, magnetic resonance imaging.
trace > 3.28 ppm (other by-products than phosphocholine), whereas the remaining 11 of 46 cases showed negative tCholine trace on both peaks. BI-RADS Classification Reading all parameters together, 58.54% (48 of 82) of lesions were assigned BI-RADS 2 class, 3.66% (3 of 82) were assigned BI-RADS 3, which was in concordance with the histopathological results of fat necrosis, whereas the remaining 37.80% (31 of 82) were assigned BI-RADS 4 and 5, which matched the histopathological results of cancerous lesions. It has to be stressed on that diffusion and spectroscopic analysis are not established parameters of the BI-RADS system yet, and lesions were assigned according to the morphology and kinetic analysis into the different BI-RADS classes; still, diffusion and spectroscopic analysis endorsed the final selection of the BI-RADS class without the need to overclassify lesions. Radiologic-Pathologic Correlation of Cases with Fat Necrosis
Morphology and Signal Intensity A fat signal was reproducible in 98% (50 of 51) of the lesions (Table 1, Fig 1), the remaining case lacking fat signal was replaced by signal voids to be pathologically proven hemorrhage on a background of florid fibrosis (Fig 2). Enhancement Characteristics An enhancement was encountered only in 2 of 51 cases (4%), which was heterogeneous, still exhibiting type I curve of progressive enhancement (Table 1). Histopathologically, this is explained by the coexisting different stages of inflammatory changes and fibrosis (Fig 2). Nearly three quarter or 74.5% (38 of 51) of the cases revealed enhancement of the surrounding fibrosis, again of type I, none showed type II or III curve patterns, whereas the remaining 21.5% (11 of 51) revealed no enhancement of the surrounding fibrous reac4
tion, pathologically proven to be mature fibrous tissue, including the lesion lacking fat signal (Fig 1). DWI and Spectroscopic Analysis None of the lesions of fat necrosis revealed diffusion restriction, even lesions showing heterogeneous enhancement had ADC values of 1.6 and 1.9 × 10−3 mm2/s, respectively (Table 1, Figs 1, 2). Spectroscopic analysis was performed to 15 of 51 cases of fat necrosis; 2 heterogeneously enhancing lesions and 13 cases with extended surrounding enhancing fibrosis. None revealed choline trace at 3.22 ppm, 4 of 15 (26.67%) displayed positive tCholine trace > 3.28 ppm, including one of the heterogeneously enhancing lesions at 3.34 ppm, whereas the remaining 11 of 15 (73.33%) cases showed negative tCholine trace, including the second heterogeneously enhancing lesion (Fig 2). Diagnostic Performance
To label a lesion as fat necrosis on MRI analysis, detection of fat signal in a lesion revealed sensitivity of 98.04%, specificity of 100%, positive predictive value (PPV) of 100%, and negative predictive value (NPP) of 96.88%. Nonenhancement of the lesion itself revealed sensitivity of 96.08%, specificity of 100%, PPV of 100%, and NPV of 93.94%. However, adding the nonrestriction on diffusion analysis increased the sensitivity and specificity to 100% and PPV of 100% for fat necrosis and hence an NPV for malignancy of 100%. When putting spectroscopic analysis to the test, the lack of tCholine at 3.22 ppm yielded reproducible results as in diffusion analysis, namely, sensitivity, specificity, and PPV of 100% for fat necrosis, and in return a NPV of 100% for malignancy. DISCUSSION The appearance of fat necrosis is variable according to the fat content and the degree of inflammation and fibrosis of the lesions (1–5). MRI is considered as a problem-solving tool as it depends on detection of the molecular composition of the lesion,
Academic Radiology, Vol ■, No ■■, ■■ 2018
MRI CHARACTERISTICS OF FAT NECROSI BREAST
Figure 1. A 44-year-old female patient with a history of right breast cancer managed surgically followed by chemo- and radiotherapy on periodic follow-up. Ultrasound examination (a) identified a linear hypoechogenicity matching with postoperative scar; 2 years later, follow-up ultrasound (b) revealed irregular stellate hypoechoic lesion, raising the suspicion of malignancy. Axial T1 non–fat-suppressed (c) and T2 fat-suppressed (d) images revealed marked focal fibrosis along the operative scar associated with few internal fat foci reflecting end-stage fat necrosis with maturing fibrous tissue. Postcontrast T1 subtracted (e) and MIP (f) images show negligible enhancement of the fibrosis with central nonenhancing fat foci confirming fibrosis maturation. Diffusion-weighted image (g) with b-value of 800 showed no evidence of diffusion restriction confirming the nonrestricted values on ADC map (h), endorsing the BI-RADS 2 class assigned to the lesion based on morphology and kinetics analysis and excluding suspicion of recurrence.
facilitating the depiction of fat, which is still the most reliable imaging sign for the diagnosis of fat necrosis (2,7). Still, the current literature (2,7) solely describes the various morphologic changes of fat necrosis. Unfortunately, there will be instances in which fat necrosis is indistinguishable from cancer on MRI. Lesions with no gross lipid and a preponderance of fibrosis can appear spiculated and nearly solid, suggesting cancer rather than fat necrosis (2,7).
Hence, we came up with the idea of putting MRI of the breast to the test by reviewing not only the morphologic but also the new sequences that depend on the biological behavior of the lesion such as DWI and spectroscopic analysis in an attempt to use it to differentiate between breast cancer and fat necrosis. All lesions in the current study were nonclassical for fat necrosis on both mammography and sonography, this might be 5
HASSAN ET AL
Academic Radiology, Vol ■, No ■■, ■■ 2018
Figure 2. A 54-year-old female patient with a history of right breast cancer managed surgically followed by chemo- and radiotherapy on follow-up. Mammographic MLO view of the right breast (a) revealed upper outer quadrant focal asymmetry, and ultrasound (b) identified a suspicious focus within the operative bed. Axial T1 non–fatsuppressed (c) and T2 fat-suppressed (d) images revealed a markedly hypointense focus on both sequences along the operative scar associated with surrounding edema. No fat signal is identified, still marked signal void of hemosiderin deposition is the most likely cause. Postcontrast subtracted image for mean curve analysis (e) revealed focal enhancement of type I b progressive enhancement with delayed washout. No central nonenhancement identified, making the diagnosis equivocal for recurrence; still no choline trace was identified on spectroscopic analysis (f), with the value of tCholine being zero at the 3.22 ppm point along the scale, endorsing the BIRADS 3 class assigned to the lesion based on morphology and kinetics analysis and excluding the suspicion of recurrence. Diffusion-weighted image (g) with b-value of 800 and ADC map (h) showed no evidence of diffusion restriction, confirming the spectroscopic analysis and excluding recurrence. Low power microscopic examination of the lesion (i) shows multiple communicating fat cysts, hemorrhage, and a diffuse mixed inflammatory infiltrate (hematoxylin and eosin [H&E] ×100), higher power examination (j) shows large fat cyst surrounded by necrosis entangling karyorrhectic debris and a mixed inflammatory infiltrate (H&E × 200). High power examination (k) confirms the inflammatory infiltrate formed by large foamy histiocytes (arrows), neutrophils, lymphocytes, and hemosiderinladen macrophages (H&E × 400).
attributed to the fact that the fibrosis was taking the upper hand is so florid, even experienced radiologists could not confidently diagnose it as fat necrosis especially in the setting of the postmanaged breast. Of 82 examined patients, 51 were pathologically proven to be fat necrosis. This was concordant with the final MRIBI-RADS category based on the signal intensity, enhancement characteristic, DWI, and spectroscopic analysis. Regarding the signal intensity of the lesions, 98% (50 of 51) of the pathologically proven fat necrosis lesions detected in this series did show fat signal, which concurs with current literature (1–7,17,18), sometimes termed black hole as they become markedly hypointense compared to the surrounding fat (18). 6
The only case not conforming to this finding, which can be labeled lipid poor, is explained in the literature (1,3–7) by the fact that lack of gross lipid and preponderance of fibrosis within a lesion gives it the appearance of nearly solid, suggesting cancer rather than fat necrosis. Moreover, hemorrhage, which is a described component of the inflammatory changes taking place in fat necrosis, can lead to signal voids on MRI due to its iron content (3,4), both phenomena were proven present during the histopathological analysis of this case. The current literature (1,3–7) finds kinetic analysis to be of little help, as fat necrosis exhibits the full spectrum of benign and malignant enhancement patterns as a result of a balance between the degree of inflammation and maturing fibrosis. This was only observed in 4% (2 of 51) of the cases. In contrast,
Academic Radiology, Vol ■, No ■■, ■■ 2018
MRI CHARACTERISTICS OF FAT NECROSI BREAST
Figure 2. (continued).
96% (49 of 51) of the cases did show nonenhancement of the core of the lesion containing gross fat foci, whereas the surrounding fibrosis enhancement varied according to degree of maturity; this yielded a PPV of 100% for a lesion to be fat necrosis, which is considered as an assuring point (1,3–7,17). The existing studies about fat necrosis appearance by MRI did not address diffusion restriction as sole research. However, DWI has been examined in the context of differentiating benign from malignant breast lesions; a number of these studies (8,10,11,15,19) did show that fat necrosis does not express diffusion restriction. This concurred with the current findings, where none of the lesions showed diffusion restriction. The spectroscopic analysis dedicated to the recognition of fat necrosis is sparse in the literature (8,19); still, the protocol of this study speculated that none of the fat necrosis cases would exhibit choline trace at 3.22 ppm, which concurred with the results. Out of the examined 15 cases, only 26.67%
(4 of 15) displayed tCholine trace > 3.28 ppm explainable as by-products of inflammatory changes, whereas 73.33% (11 of 15) revealed negative tCholine trace at both 3.22 and >3.28 ppm. The remaining 36 cases of fat necrosis were not subjected to spectroscopic analysis due to inadequate sample size or lack of enhancement of the lesion itself. Deducted from the current study, to reach a definitive diagnosis, an algorithm for review of the sequences at hand should be taken into consideration to try and reach out the maximum benefit of the examination. The typical appearance should contain fat signal, which usually exhibits “black hole effect.” Hemorrhagic component appearing as signal voids can be present. Enhancement pattern, if present, is typically rim enhancement of the surrounding fibrosis regardless of the type of curve pattern, although type I would be a plus as fibrosis does enhance late. Sometimes the enhancing rim would be thick due to florid fibrosis giving appearance of heterogeneous 7
HASSAN ET AL
enhancement; the same situation is faced when dealing with active inflammatory process. Hence, reverting to the other sequences, no diffusion restriction and negative at 3.22 ppm or low level of tCholine trace > 3.28 ppm would be correspondingly in favor of fat necrosis and inflammatory changes. The limitation of our study was that most cases positive for fat necrosis were already managed cases; therefore, the probability of postmanagement sequel is undeniably higher than the rate of recurrence. Conversely, this point could be used as an advantage to enroll MRI of the breast as part of the routine examination of challenging managed breast. This is due to that fact that we are dealing with a probable recurrence rate that might be as high as 2.5% per year (17), which is problematic to managed cases of breast cancer, particularly those refusing to undergo biopsy or have a contraindication to biopsy as it will yield desirable results with less invasive maneuver. A second limitation is that there is still no standardized cut-off value along the gray zone of ADC values of 1–1.5 × 10−3 mm2/s to label a lesion as benign or malignant. However, based on several research articles (15,16), the suggested cut-off values based on ROC analysis yielding up to 90% sensitivity and 80% specificity ranged from 1.23 to 1.26, hence we decided to use a slightly higher value of 1.3. Needless to say, one cannot depend solely on DWI as a differentiating point for lesion’s nature and it has to be still read as just a part of the whole study and not as a stand-alone sequence. Additional limitation includes the fact that spectroscopic analysis was done only to cases with enhancing lesions, although this is the standard of care, it consequently resulted in a small sample size of fat necrosis lesions tested. Hence, we propose further studies to be carried out even in lack of enhancement to detect any signal in the surrounding inflammatory reaction to test if end-stage fat necrosis—fibrosis—does in fact lack definite by-products. In conclusion, MRI proved to be of value in differentiating fat necrosis from malignancy based on the molecular composition of fat necrosis, which is clearly depicted by MRI. Moreover, multisequence approach proved to be of value in cases where the classical fat intensity signal could not be detected. Hence, in cases showing the previously described MRI signs of fat necrosis, the need for invasive confirmation by biopsy might not be required.
8
Academic Radiology, Vol ■, No ■■, ■■ 2018
REFERENCES 1. Tan PH, Lai LM, Carrington EV, et al. Fat necrosis of the breast—a review. The Breast 2006; 15:313–318. 2. Chan LP, Gee R, Keogh C, et al. Imaging features of fat necrosis. AJR Am J Roentgenol 2003; 181:955–959. 3. Daly CP, Jaeger B, Sill DS. Variable appearances of fat necrosis on breast MRI. AJR Am J Roentgenol 2008; 191:1374–1380. 4. Taboada JL, Stephens TW, Krishnamurthy S, et al. The many faces of fat necrosis in the breast. AJR Am J Roentgenol 2009; 192:815–825. 5. Kerridge WD, Kryvenko ON, Thompson A, et al. Fat necrosis of the breast: a pictorial review of the mammographic, ultrasound, CT, and MRI findings with histopathologic correlation. Radiol Res Pract. 2015; 2015:613139. doi:10.1155/2015/613139. 6. Chala LF, de Barros N, Moraes P, et al. Fat necrosis of the breast: mammographic, sonographic, computed tomography, and magnetic resonance imaging findings. Curr Probl Diagn Radiol 2004; 33:106–126. 7. Sergi Ganau S, Tortajada L, Escribano F, et al. The great mimicker: fat necrosis of the breast—magnetic resonance mammography approach. Curr Probl Diagn Radiol 2009; 38:189–197. 8. Stanwell P, Gluch L, Clark D, et al. Specificity of choline metabolites for in vivo diagnosis of breast cancer using 1H MRS at 1.5 T. Eur Radiol 2005; 15:1037–1043. 9. Sardanelli F, Fausto A, DiLeo G, et al. In vivo proton MR spectroscopy of the breast using the total choline peak integral as a marker of malignancy. AJR Am J Roentgenol 2009; 192:1608–1617. 10. Partridge SC, DeMartini WB, Kurland BF, et al. Quantitative diffusionweighted imaging as an adjunct to conventional breast MRI for improved positive predictive value. AJR Am J Roentgenol 2009; 193:1716–1722. 11. Kul S, Cansu A, Alhan E, et al. Contribution of diffusion weighted imaging to dynamic contrast-enhanced MRI in the characterization of breast tumors. AJR Am J Roentgenol 2011; 196:210–217. 12. Gonzales MA, Castañon AI, Baltar N, et al. Diffusion weighted MR imaging assessment after breast conservative treatment. Poster session in ECR 2012 Educational Exhibit. March 1–5, 2012; Vienna, Austria. doi:10.1594/ecr2012/C-1065. 13. Şahin C, Arıbal E. The role of apparent diffusion coefficient values in the differential diagnosis of breast lesions in diffusion-weighted MRI. Diagn Interv Radiol 2013; 19:457–462. 14. Hassan HHM, Zahran MHM, Hassan HE, et al. Diffusion magnetic resonance imaging of breast lesions: initial experience at Alexandria University. Alexandria J Med 2013; 49:2652–2672. 15. Hetta W. Role of diffusion weighted images combined with breast MRI in improving the detection and differentiation of breast lesions. Egypt J Radio Nuc Med 2015; 46:259–270. 16. Tsushima Y, Takahashi-Taketomi A, Endo K. Magnetic resonance (MR) differential diagnosis of breast tumors using apparent diffusion coefficient (ADC) on 1.5-T. J Magn Reson Imaging 2009; 30:249–255. 17. Chansakul T, Lai KC, Slanetz PJ. The post-conservation breast: part 2, imaging findings of tumor recurrence and other long-term sequelae. AJR Am J Roentgenol 2012; 98:331–333. 18. Trimboli RM, Carbonaro LA, Cartia F, et al. MRI of fat necrosis of the breast: the “black hole” sign at short tau inversion recovery. Eur J Radiol 2012; 81:e573–e579. 19. Sardanelli F, Carbonaro LA, Montemezzi S, et al. Clinical breast MR using MRS or DWI: who is the winner? Front Oncol 2016; 6:e217.