Eur Radiol (2008) 18: 2370–2380DOI 10.1007/s00330-008-1029-3 BREAST

Diane M. RenzPascal A. T. BaltzerJoachim BöttcherFady ThaherMieczyslaw GajdaOumar CamaraIngo B. RunnebaumWerner A. Kaiser

Received: 27 September 2007Revised: 10 March 2008Accepted: 30 March 2008Published online: 4 June 2008# European Society of Radiology 2008

Magnetic resonance imaging of inflammatorybreast carcinoma and acute mastitis.A comparative study

Abstract The aim of this study was toevaluate the potential of magneticresonance mammography (MRM) todistinguish inflammatory breastcarcinomas (IBC) from acute mastitis(AM). This study compared MRMexaminations of two selected groupsof patients: 48 subjects with IBC and42 patients with AM. No statisticaldifferences were revealed between thetwo groups for morphology of massesand of non-mass-like enhancement,breast enlargement, diffuse skinthickening, abnormal nippleconfiguration, prominent vessels, andalso for cutaneous/subcutaneous,perimamillar and diffuse oedema.However, initial and postinitialdynamic characteristics significantlydiffered between the two groups(p<0.001). In IBC, more masses witha greater average size were detected(p<0.05). The followingmorphological criteria were alsoobserved more often in IBC (p<0.05):T2-hypointensity of masses(77.5%/18.2%), blooming sign(62.5%/31.8%), infiltration ofpectoralis major muscle (interruptionof fat plane: 54.2%/16.7%; pathologi-

cal enhancement: 33.3%/7.1%),perifocal (66.7%/33.3%), prepectoral(72.9%/31.0%) and intramuscular pec-toral oedema (41.7%/7.1%). The mainlocalisation of AM was subareolar, ofIBC central or dorsal (p<0.001). Thediscrimination between AM and IBCremains a diagnostic challenge becauseof overlapping imaging features. How-ever, the combination of multipledynamic and morphological MRMcriteria seems to have the potential for adifferential diagnosis.

Keywords Inflammatory breastcarcinoma . Acute mastitis .Breast infection . MRmammography . Morphology .Dynamics

Abbreviations IBC: inflammatorybreast carcinomas . AM: acutemastitis . MRM: magnetic resonancemammography . BI-RADS: breastimaging reporting and data system .ROI: region of interest

Introduction

Breast infection can occur in healthy, non-lactating womenof all ages [1]. Currently, there seems to be an increasingrate of benign and malignant inflammation in the non-lactating breast [2, 3]. It remains a challenge to distinguishacute mastitis (AM) from malignancy, especially frominflammatory breast carcinomas (IBC), by clinical or

imaging features [1, 4, 5, 6]. IBC account for about 1–4% of all breast cancer cases with a tendency to metastasiseat an early stage, resulting in a poor prognosis [3, 7].

In IBC as well as AM, typical clinical manifestations areenlargement, tenderness, increased cutaneous temperatureand dermal redness of the involved breast [1, 2, 4, 7].Despite progress in techniques and experience, X-raymammography and ultrasonography are not able to

D. M. Renz . P. A. T. Baltzer .J. Böttcher . F. Thaher . W. A. KaiserInstitute of Diagnosticand Interventional Radiology,Friedrich Schiller University Jena,Erlanger Allee 101,07740 Jena, Germany

M. GajdaInstitute of Pathology,Friedrich Schiller University Jena,Ziegmühlenweg 1,07740 Jena, Germany

O. Camara . I. B. RunnebaumClinic of Gynaecology,Friedrich Schiller University Jena,Bachstr. 18,07740 Jena, Germany

D. M. Renz (*)Institute of Diagnosticand Interventional Radiology,Gynaecological Radiology,Friedrich Schiller University Jena,Erlanger Allee 101,07740 Jena, Germanye-mail: diane.renz@med.uni-jena.deTel.: +49-3641-9324849Fax: +49-3641-9324832

precisely differentiate between IBC and AM. Both diseasespresent non-specific imaging features such as a focal ordiffuse asymmetric density in X-ray mammography [1, 4,5]; in ultrasonography, hypoechoic focal masses werefrequently found in IBC as well as in AM [1, 4, 5]. Theseoverlapping clinical and imaging features can lead to adelay in the diagnosis of IBC [3, 6, 7]. Because of theserious consequences of this hesitation, a histologicalverification of IBC by needle or punch skin biopsy isessential if any diagnostic doubt exists [8, 9].

Magnetic resonance mammography (MRM) has becomea well-established method in the diagnosis and staging ofinvasive breast carcinoma [10, 11]. MRM has the highestcurrently reported sensitivity of all imaging modalities inbreast cancer detection, approaching 99% with varyingobserver-dependent specificity levels [10, 12, 13, 14].Dynamic characteristics, such as plateau phenomenon andwash-out sign, have been proven in several publications tobe an important criterion for the detection of invasivebreast carcinomas [12, 13, 14, 15]. Only a few reportsparticularly considered T2-weighted sequences to evaluatespecific morphological criteria, such as tumour-associatedoedema [16, 17, 18]. Based on dynamic and morphologicalcharacteristics, MRM has proven benefit in preciselydiscriminating between malignant and benign lesions [18,19].

The present knowledge about the typical appearance ofIBC and AM in MRM is fragmentary, caused by limiteddata. Some authors conclude that MRM is not able todistinguish between the two inflammation entities [8, 9,20, 21]. However, all previous reports were performed onvery small samples, with a maximum of 15 patients foreach group, and evaluated only a limited quantity ofimaging parameters. Most of these analysed MRMcriteria, including skin thickening, breast enlargementand enhancement pattern of the infiltration, showedsimilar prevalence in IBC and AM [8, 9, 20]. The aim ofthis study was to evaluate the potential of MRM todifferentiate between IBC and AM, considering a multi-tude of dynamic and morphological parameters in T1-and T2-weighted sequences.

Materials and methods

Subjects

All evaluated MRM examinations were performed at ourradiological institute within four years. To obtain a reliableevaluation and comparable imaging findings, all factorsthat could potentially influence dynamic or morphologicalfeatures were considered. These extended demandsresulted in the exclusion of patients having receivedchemotherapy, breast conservation, mastectomy or radia-tion therapy prior to MRM. No subject had undergonebiopsy of the affected breast in the past 3 months. Because

of the effects of hormone replacement therapy on non-mass-like enhancements, patients were only enrolled in thisinvestigation if they have not received hormone replace-ment therapy within the past 6 weeks. Premenopausalpatients were only considered if the MRM examinationwas performed in the first half of their menstrual cycle.Forty-eight patients with IBC and 42 subjects with AMfulfilled the detailed inclusion criteria. All patients withAM suffered from non-puerperal acute infectious mastitis.All patients gave their consent for scientific use of theirclinical and radiological data.

The ages of the 48 patients with IBC ranged from 36 to83 years [mean=56.8, SD (standard deviation)=12.8]. Theaverage age of the 42 subjects with AM was significantlylower (p<0.01): mean=48.4 (SD=10.9) with a range from25 to 72 years. Prior to the MRM examination, the averageduration of the clinical symptoms of the IBC patients was20.2 days (SD=8.8 days) with a variation from 6 to52 days. In comparison, the subjects with AM had hadclinical symptoms for a mean duration of 11.7 days (SD=6.7 days) before MRM, ranging from 5 to 28 days (p<0.001).

MR mammography

A predefined technical protocol of MRM examination wasused. All images were obtained by a 1.5-T scanner(Symphony; Siemens Medical Solutions, Erlangen, Ger-many) using a double breast coil with the patient in a proneposition. For the dynamic study, multi-slice 2D FLASH(fast low-angle shot) T1-weighted sequence images wereacquired with the following parameters: repetition time113 ms; echo time 4.6 ms; matrix 384×384; flip angle 80°;slice thickness 3.0 mm; field of view 350 mm; axialorientation. Axial T2-weighted turbo-spin-echo images(repetition time 8,900 ms; echo time 207 ms; matrix 512×512; flip angle 90°; field of view 350 mm) were obtained inidentical slice positions. After acquisition of a native T1-weighted sequence, 0.1 mmol/kg body weight gadopente-tate dimeglumine (Magnevist; Bayer Schering Pharma,Berlin, Germany) was intravenously administered as arapid bolus within 10 s followed by 20 ml saline flush; 30 safter bolus injection and saline administration, dynamicMRM was continued with the same sequence parametersand under identical tuning conditions at 1-min intervals upto 7 min after contrast media administration. The nativeunenhanced images were subtracted from post-contrastdynamic images.

Evaluation criteria

All images were retrospectively analysed by two radiolo-gists specialised in MRM. The observers were blinded tothe diagnosis and the histopathological findings. The two

2371

radiologists evaluated independently, blinded to each otherand under the same conditions. In case of a consensusbetween the two observers, their findings were considered.When discrepancies between the first two readers occurred,a third radiologist with detailed MRM experience made thefinal decision after a thorough imaging analysis and adialogue with the two other observers.

Computer software (Syngo; Siemens Medical Solutions,Erlangen, Germany) was used to ascertain signal intensitytime curves in areas of kinetic increase based on contrastenhancement. Manual ROIs (regions of interest, 3×3pixels) were placed by the observers focussing on the mostenhancing part of region. If the lesion presented masses,three ROIs were placed in each mass. The contrastenhancement was then assessed as the mean signalintensity increase of all masses. In absence of masses, thesignal characteristic of the lesion was measured by threeROIs in the most enhancing region of the non-mass-likeenhancement.

For the evaluation of the images, the following dynamicand morphologic criteria were recorded:I. Dynamic criteria [14, 15, 18, 19]

a. Initial signal increase. The initial intensity increaseof the lesion was measured within the first 90 s afterbolus injection in the following categories: (1)<50%; (2) 50–100%; (3) >100%.

b. Postinitial contrast enhancement was assessed at 3to 7 min after contrast media administration. Threetypes of postinitial curves were defined: (1) contin-uous increase (increase >10%); (2) plateau phe-nomenon (deviation of the signal curve between+10% and–10%); (3) wash-out sign (decrease>10%).

II. Analysis of masses [16, 18, 22, 23]a. Shape. According to the Breast Imaging Reporting

and Data System (BI-RADS) atlas, the configura-tion observed in the majority of all detected masseswas recorded as follows: (1) round; (2) oval; (3)lobulated; (4) irregular.

b. Margin. Based on BI-RADS, the margin of themajority of all detected masses was described as (1)smooth; (2) irregular; (3) spiculated.

c. Blooming phenomenonmeans that 1 min after bolusinjection a fast enhancing lesion shows a sharplyconfigurated border; the lesion’s rim becomes moreand more unsharp up to 7 min after contrast mediaadministration.

d. Signal intensity in T2-weighted images. Malignanttumours are supposed to be mostly hypointense inT2-weighted sequences, whereas benign lesionsoften show hyperintense or isointense signal inten-sity compared to normal breast tissue. The follow-ing signal intensity of the detected masses wasdetermined: (1) hyperintense or isointense; (2)hypointense.

III. Evaluation of non-mass-like enhancement [22]a. Spatial distribution. The spatial distribution of the

non-mass-like enhancement was ascertained in thefollowing categories, according to the BI-RADSatlas: (1) linear/ductal; (2) segmental; (3) regional;(4) multiple regions; (5) diffuse.

b. Internal enhancement pattern. The internal mor-phology of the enhancement was described asfollows: (1) stippled/punctate; (2) clumped; (3)dendritic; (4) non-specific.

c. Main localisation. The main localisation of the non-mass-like enhancement was divided into three partsof the affected breast: (1) subareolar; (2) central; (3)dorsal.

IV. Analysis of breast parameters [8, 9, 17, 18, 24, 25, 26, 27]a. Breast enlargement. In inflammatory carcinomas as

well as in mastitis, an enlargement of the affectedbreast, in comparison with the contralateral one, isfrequently found. The existence or absence of thisparameter was recorded.

b. Diffuse skin thickening. If the skin of the affectedbreast measures >3 mm and is broader than thecontralateral one, this observation can be interpretedas a sign of skin involvement. The presence orabsence of this criterion was assessed using nativeT1-weighted images.

c. Punched-out sign. The definition of this phenom-enon is as follows: 1 min after bolus injection, someareas of the skin show fast, accentuated enhance-ment. These focal parts may represent a tumouralskin infiltration, whereas the surrounding skindemonstrates a slow-continuous signal increase upto 7 min after contrast media administration. Theexistence or absence of this sign was considered.

d. Analysis of nipple. The nipple was considered asnormal if it presented superficial linear enhance-ment and symmetrical morphology compared to thecontralateral breast. The nipple was recorded asabnormal in case of a nodular enhancement, anincreased thickening compared to the contralateralone and an interruption of the mamillary edge aftercontrast media administration.

e. Evaluation of pectoralis major muscle. It wasdetermined if the fat plane of the pectoralis majormuscle was interrupted or not. Furthermore, theenhancement of the pectoralis major muscle wasconsidered, whether the muscle presented asym-metric strong (pathological) or symmetric baselinesignal increase after contrast media administration.

f. Prominent vessels. It was evaluated if the vessels ofthe affected breast were prominent in diameter andquantity compared with the contralateral breast. Thepresence or absence of this sign was recorded.

g. Oedema. The following categories of differentoedema localisations were evaluated in the T2-

2372

weighted images: (1) cutaneous/subcutaneous;(2) perimamillar; (3) perifocal; (4) diffuse; (5)prepectoral; (6) intramuscular pectoral.

h. Lymph node involvement. Axillary lymph nodes wereconsidered as malignant if they presented a size>1.5 cm and a round shape [short-to-long axis ratio (S/L) ≥0.5] without hilar fat. The nodes were determinedas benign reactive if they demonstrated a size >1.5 cm,an oval configuration (S/L<0.5) and hilar fat.

Statistical analysis

The prevalence of all evaluation criteria was separatelyconsidered for the sample of 48 IBC patients and thegroup of 42 subjects with AM. For each dynamic andmorphological parameter, the prevalence was comparedbetween the two groups. It was assumed as the nullhypothesis that the two cohorts do not present differentMRM characteristics.

The statistical analysis was performed using SPSSversion 15.0 for Windows (SPSS; Chicago, IL, USA). TheFisher’s exact test was used to examine the significance ofthe association between two variables in a 2×2 con-tingency table. Evaluating variables with more than twocategories, the Pearson’s chi-square test was performed.The Student’s t-test was used to examine significantdifferences between means of two independent samples. Ap-value less than 0.05 (two-sided) was considered toindicate a significant difference.

Results

Clinical and additional imaging characteristics

The groups of IBC and AM present similar clinicalsymptoms (for details see Table 1), except for theexistence of a palpable mass. X-ray mammography didnot reveal any significant differences between the twodiseases regarding diffuse increased breast density andsuspicious microcalcifications. In ultrasonography, IBCand AM present parenchymal irregularities in a similarprevalence. However, masses were detected morefrequently in IBC in X-ray mammography and ultra-sonography (p<0.01). In two patients with IBC (4.2%)and seven subjects suffering from AM (16.7%), X-raymammography did not present any abnormalities. Allof these patients had very dense breast tissue.

Histopathological findings and treatment

In all IBCs, the diagnosis of invasive breast carcinoma washistopathologically confirmed after the MRM examination.

By histology, 35 of 48 IBC (72.9%) were categorised asinvasive ductal; the remaining 13 cases (27.1%) weredefined as invasive lobular. Among the 48 IBC, 16 (33.3%)were verified as G2 (moderately differentiated) and 32(66.7%) as G3 (poorly differentiated). None of the inflam-matory carcinomas was defined as well-differentiated (G1).

Thirty-two of 48 IBC (66.7%) showed lymphaticinfiltration of the skin at histology. The remaining 16tumours (33.3%) were characterised by the appearance ofclinical signs and the absence of histopathologically provenlymphatic infiltration of the skin. Our data did not showsignificant differences between the existence or absence ofhistologically verified lymphatic invasion of the skin andthe appearance of almost all evaluated dynamic andmorphological criteria. Only the two parameters of skininvolvement, i.e., diffuse skin thickening and punched-outsign, were detected more often in IBC with histologicallyproven lymphatic infiltration.

All 42 patients with AM were treated with oralantibiotics. In all subjects, the clinical symptoms subsidedrapidly during antibiotic treatment. The diagnosis wasmade by histological proof of acute suppurative inflam-mation in 24 of 42 AM patients (57.1%), either by biopsy(n=16; 38.1%) or surgery with abscess drainage (n=8;19.0%). In the remaining 18 subjects (42.9%), a follow-upwas performed by MRM documenting complete remissionof pathological imaging findings within months.

After histological verification of breast cancer, most ofour IBC patients (n=34; 70.8%) underwent neoadjuvant

Table 1 Clinical, X-ray mammographical and ultrasonographicfeatures of inflammatory breast carcinomas and acute mastitis

Inflammatorybreastcarcinomas(n=48)

Acutemastitis(n=42)

P-valuesof Fisher’sexact test(two-sided)

Clinical symptomsErythema 44 (91.7) 40 (95.2) n.s.*Increased cutaneoustemperature

29 (60.4) 33 (78.6) n.s.*

Breast pain 31 (64.6) 35 (83.3) n.s.*Palpable mass 33 (68.8) 16 (38.1) <0.01

X-ray mammographical featuresDiffuse increasedbreast density

41 (85.4) 31 (73.8) n.s.*

Suspiciousmicrocalcifications

9 (18.8) 2 (4.8) n.s.*

Existence of masses 32 (66.7) 15 (35.7) <0.01No abnormalities detected 2 (4.2) 7 (16.7) n.s.*

Ultrasonographic featuresParenchymal irregularities 41 (85.4) 32 (76.2) n.s.*Existence of masses 36 (75.0) 19 (45.2) <0.01No abnormalities detected 3 (6.3) 7 (16.7) n.s.*

Numbers in parentheses are percentages; *no significant difference

2373

chemotherapy and subsequent breast surgery. Eleven of 48subjects with IBC (22.9%) had initial mastectomycombined with adjuvant chemotherapy. In view of theirpoor physical condition, chemotherapy alone was per-formed in three patients with IBC (6.3%).

Dynamic criteria

Signal characteristics showed significant differences be-tween the two groups regarding initial signal increase andpostinitial enhancement (Table 2). Forty-one of 48 IBC(85.4%) and 19 of 42 AM (45.2%) presented initial signalintensity >100%; 68.8% of IBC (n=33 of 48) and 14.2%(n=6) of 42 AM showed a wash-out sign (Table 2).

Analysis of masses

Eight IBC (16.7%) and 20 AM (47.6%) were exhibiteddiffuse changes without existence of a mass (Fig. 1). In the

remaining 40 IBC, a larger number of masses was detected(mean=5.4, SD=4.6; Fig. 1) compared with the 22 AMshowing masses (mean=3.1, SD=3.8; p<0.05). Theaverage size of the masses was also significantly greaterin IBC (mean=2.4 cm, SD=1.2 cm) compared with AM(mean=1.2 cm, SD=0.6 cm; p<0.001). Summarising,75.0% of all 48 IBC (n=36) and 31.0% of all 42 AM (n=13) presented masses >1 cm (p<0.001).

Based on the BI-RADS criteria, the shape and margin ofthe detected masses did not differ between the two groups(Table 3). Blooming phenomenon was detected in 25 of 40IBC (62.5%) and in 7 of 22AM (31.8%; p<0.05). In 31 of 40IBC (77.5%) and in 4 of 22 AM (18.2%), the majority ofmasses were hypointense in T2-weighted images (p<0.001).

Evaluation of non-mass-like enhancement

In all evaluated subjects, a non-mass-like enhancement wasdetected (Fig. 1). In case of the existence of masses, thistype of contrast enhancement was observed adjacent to the

Table 2 Signal kinetic characteristics of inflammatory breast carcinomas and acute mastitis

Inflammatory breastcarcinomas (n=48)

Acute mastitis (n=42) P-values of Pearson’schi-square test (two-sided)

Initial signal increase<50% 1 (2.1) 5 (11.9)

9=

;< 0:00150–100% 6 (12.5) 18 (42.9)

>100% 41 (85.4) 19 (45.2)Postinitial enhancementContinuous increase 4 (8.3) 18 (42.9)

9=

;< 0:001Plateau phenomenon 11 (22.9) 18 (42.9)

Wash-out sign 33 (68.8) 6 (14.2)

Numbers in parentheses are percentages

Fig. 1 The subtraction images1 min after bolus injectiondemonstrate an acute mastitis(a) and an inflammatory carci-noma (b; invasive lobular, G3)in the right breast. Both casespresent a diffuse, stippled non-mass-like enhancement. Theinflammatory carcinoma alsoshows multiple masses (contin-uous arrows). The nipple of theacute mastitis demonstrates anodular enhancement with aninterruption of the mamillaryedge (continuous arrow). Inboth images, a diffuse skinthickening and prominent ves-sels are visualised (discontinu-ous arrows)

2374

masses. Based on the BI-RADS criteria, the spatialdistribution and the internal morphology of this enhance-ment did not differ between the two groups (Table 4).However, the main localisation of the enhancement of AMwas subareolar, of IBC central or dorsal (p<0.001).

Analysis of breast parameters

No significant differences of the following breast para-meters were detected between the two diseases: breastenlargement, diffuse skin thickening, abnormal nippleconfiguration and prominent vessels (Table 5; Fig. 1). Inthe group of IBC, the pectoralis major muscle was moreoften involved (Table 5; Fig. 2). The skin demonstratedmore frequently a punched-out sign in IBC than in AM.

All evaluated cases presented unilateral oedema in theaffected breast (Table 5; Fig. 2). Cutaneous/subcutaneous,

perimamillar and diffuse oedema was similarly observed inthe two groups. However, oedema was significantlydetected more often in IBC in the following localisations:perifocal, prepectoral and intramuscular pectoral. In 32 of48 IBC (66.7%), axillary lymph nodes were considered asmalignant. Twenty of 42 subjects with AM (47.6%)showed axillary lymph node enlargement with nodesdemonstrating a benign reactive appearance.

Focussing on IBC, cases with histologically verifiedlymphatic infiltration of the skin presented a diffusethickening and a punched-out sign of the skin significantlymore frequently than tumours without positive lymphaticinvasion at pathology. Among 32 IBC with histologicallyproven lymphatic invasion of the skin, 30 (93.8%)demonstrated a diffuse thickening and 23 (71.9%) showeda punched-out sign. In the remaining 16 IBC, a diffuse skinthickening was detected in 10 (62.5%; p<0.05), a punched-out sign was observed in 4 tumours (25.0%; p<0.01).

Table 3 Analysis of masses in inflammatory breast carcinomas and acute mastitis

Inflammatory breastcarcinomas (n=40)

Acute mastitis (n=22) P-values of Pearson’schi-square test (two-sided)

ShapeRound 15 (37.5) 10 (45.5)

9=

;n:s: �Oval 8 (20.0) 3 (13.6)

Lobulated 5 (12.5) 3 (13.6)Irregular 12 (30.0) 6 (27.3)

MarginSmooth 7 (17.5) 10 (45.5)

9=

;n:s: �Irregular 23 (57.5) 9 (40.9)

Spiculated 10 (25.0) 3 (13.6)

Numbers in parentheses are percentages; * no significant difference

Table 4 Evaluation of non-mass-like enhancement in inflammatory breast carcinomas and acute mastitis

Inflammatory breastcarcinomas (n=48)

Acute mastitis (n=42) P-values of Pearson’schi-square test (two-sided)

Spatial distributionSegmental 7 (14.6) 13 (31.0)

9=

;n:s: �Regional 9 (18.7) 9 (21.4)

Multiple regions 11 (22.9) 8 (19.0)Diffuse 21 (43.8) 12 (28.6)

Internal enhancementStippled/punctate 14 (29.2) 15 (35.7)

9=

;n:s: �Clumped 11 (22.9) 9 (21.4)

Dendritic 17 (35.4) 11 (26.2)Non-specific 6 (12.5) 7 (16.7)

Main localisationSubareolar 6 (12.5) 25 (59.5)

9=

;< 0:001Central 31 (64.6) 13 (31.0)

Dorsal 11 (22.9) 4 (9.5)

Numbers in parentheses are percentages; *no significant difference

2375

Discussion

Contrast-enhanced MRM has become valuable in thedetection, diagnosis, staging of invasive breast cancer andin discriminating malignant from benign lesions [10, 11,18, 19]. However, investigations evaluating the benefit ofMRM for the differential diagnosis of IBC and AM are stilllimited. This retrospective study gives awide overview on thistopic. The main differences between our investigation andprevious reports are the relatively high numbers of patientsand the multitude of evaluation parameters, consideringdynamic as well as morphological characteristics.

Dynamic parameters seem to have the potential todistinguish IBC from AM because of their significantlydifferent signal kinetics. However, some overlap existsbetween enhancement curve types of IBC and AM. Ourfindings are in concordance with published results: Rieberet al. [8] reported that 90% of IBC and 55% of AMdemonstrate an enhancement >100% in the first minuteafter contrast media administration. According to Tomczaket al. [20], an initial signal increase >100% can be detectedin 80% of IBC and in 43% of AM.

Buadu et al. [28] found that the quantity and thedistribution of blood microvessels play a major role in the

Table 5 Breast parameters of inflammatory breast carcinomas and acute mastitis

Inflammatory breastcarcinomas (n=48)

Acute mastitis (n=42) P-values of Fisher’sexact test (two-sided)

Breast enlargement 33 (68.8) 26 (61.9) n.s.*Skin involvementDiffuse skin thickening 40 (83.3) 28 (66.7) n.s.*Punched-out sign 27 (56.3) 3 (7.1) <0.001

Abnormal nipple configuration 32 (66.7) 22 (52.4) n.s.*Pectoralis major muscleInterruption of fat plane 26 (54.2) 7 (16.7) <0.001Pathological enhancement 16 (33.3) 3 (7.1) <0.01

Prominent vessels 41 (85.4) 30 (71.4) n.s.*Oedema typesCutaneous/subcutaneous 39 (81.3) 28 (66.7) n.s.*Perimamillar 34 (70.8) 28 (66.7) n.s.*Perifocal 32 (66.7) 14 (33.3) <0.01Diffuse 43 (89.6) 35 (83.3) n.s.*Prepectoral 35 (72.9) 13 (31.0) <0.001Intramuscular pectoral 20 (41.7) 3 (7.1) <0.001

Numbers in parentheses are percentages; *no significant difference

Fig. 2 The T2-weighted imagesreveal different oedema types ofa severe acute mastitis (a) andan inflammatory carcinoma (b;invasive ductal, G3) in the leftbreast. Both images presentdiffuse, perimamillar andcutaneous/subcutaneous oedema(continuous arrows). The in-flammatory breast carcinoma(b) additionally shows prepec-toral and intramuscular pectoraloedema (discontinuous arrow);the fat plane of the pectoralismajor muscle is interrupted

2376

initial signal increase after paramagnetic contrast admin-istration. The fast contrast enhancement is associated withneoangiogenesis [29]: The formation of new capillariesfrom preexisting blood vessels is an important componentof malignant as well as inflammatory processes. In acuteinflammatory reactions, neovascularisation is generally up-regulated for brief periods based on a relatively normalvascular architecture [29, 30]. Persistent neovascularisa-tion, however, is characteristic for malignant tumours.Because angiogenic factors are continuously activated incarcinomas, the progression of the endothelium into anetwork of mature vessels is inhibited, resulting in atangled capillary architecture [28, 29]. The higher numberand the more complex network of malignant tumourmicrovessels may be the reason for the different initialcontrast medium uptake of IBC and AM.

Besides the initial signal increase, the postinitial en-hancement also significantly differed between IBC andAM in our investigation. Most of the evaluated AMshowed a postinitial continuous signal increase or a plateauphenomenon. Analysing nine patients with mastitis,Fischer et al. [9] also found a postinitial plateau phenom-enon or a slight continous increase in all examinations. InIBC, however, wash-out time curves prevail, mainly basedon persistent neovascularisation. The wash-out phenome-non of invasive carcinomas is supposed to be caused byarteriovenous anastomoses with a consequent fast contrastmedium outflow and the large number of capillaries with ahigh blood flow rate, resulting in a rapid transport ofcontrast agent from the intravascular to the interstitialcompartment [14, 15]. This mechanism also can explainthe blooming sign [23] which was significantly morefrequent in patients with IBC.

In general, capillaries induced by angiogenesis have aless intact basal membrane, which causes a penetration of

plasma proteins, resulting in oedema [30, 31]. Vascularendothelial growth factor (VEGF) is up-regulated inmalignant as well as inflammatory processes and issupposed to act as a key cytokine in breast oedemabecause of its influence on neovascularisation andincreased permeability of these new capillaries [30, 31,32]. Furthermore, blood vessels observed in carcinomasand inflammatory reactions typically show vasodilatation,which also results in oedema [15, 30, 31]. The distinctoedema is particularly responsible for the characteristicenlargement of the affected breast [8, 20]. The typical skinthickening of IBC and AM is mainly due to cutaneous/subcutaneous oedema.

Besides these pathophysiological mechanisms, theoedema of IBC is also caused by the characteristicextensive lymphovascular tumoural infiltration [4, 7]. Inthe 48 IBC of our study, prepectoral and intramuscularpectoral oedema, which is rarely observed in noninflam-matory invasive carcinomas, was frequently detected. Thelymphatic drainage of the breast is performed anteropos-teriorly to lymph nodes in the axilla and along the internalmammary vessels [33]. This anteroposterior lymphaticdrainage might be the reason for the dorsally locatedoedema of IBC, due to infiltration by tumour emboli,which obstruct the lymphatic vessels.

The characteristic lymphovascular infiltration of IBC issupposed to also cause the skin thickening [7]. Besidesthickening, another imaging characteristic of IBC seems tobe the pathological skin contrast enhancement. In seven ofnine inflammatory carcinomas, Lee et al. observed abnor-mal, strong signal increase of the skin in MRM [34]. Theauthors concluded that pathological enhancement of thethickened skin is useful in the diagnosis of IBC. A reliableindicator for the lymphatic infiltration of the skin may bethe punched-out sign [24]. In our study, IBC with histo-

Fig. 3 A 42-year-old womanwith histologically proven acutemastitis involving the rightbreast. The subtraction image1 min after bolus injection (a)visualises a mass with a rim-likesignal increase (arrow) and asurrounded stippled enhance-ment. In the T2-weighted image(b), the abscess formation ishyperintense (arrow)

2377

logically proven lymphatic skin infiltration presented thisphenomenon significantly more frequently than tumourswithout lymphatic tumoural involvement at pathology.

Based on the BI-RADS classification the morphology ofmasses was similar between the two diseases. However, thesignal intensity of the masses in the T2-weighted sequencesdiffered. Most detected masses in benign breast infectionsare abscesses, with a characteristic high signal intensity inT2-weighted images and a rim-like contrast enhancement(Fig. 3), or suppurative granulation tissue during theformation of abscesses [1, 2, 8, 35]. In contrast, fibrotictissue as a consequence of desmoplastic reaction isaccentuately verified in invasive carcinomas leading to ahypointensity in T2-weighted sequences [16, 18].

The main localisation of the non-mass-like enhancementof IBC seems to be central or dorsal, in concordance withthe dorsally occuring oedema with an infiltration of thepectoralis major muscle in many cases. The main localisa-tion of acute infectious inflammation, however, is mostlysubareolar caused by the ascending infection from thenipple by bacteria, such as staphylococcae [1, 2]. Aninfiltration of the mammary gland by macrophages andneutrophil granulocytes can be revealed at histopathology(Fig. 4). Typical histological findings of IBC are tumouremboli in lymphatic capillaries and infiltration of thestroma by tumour cells (Fig. 4).

Mastitis is not a single specific histopathological entity,but comprises several inflammatory diseases that can occurin the lactating and not-lactating breast [1, 2, 5, 6, 35]. Non-lactational mastitis includes many different subtypes, suchas plasma cell mastitis, a non-bacterial inflammation inolder women due to extravasation of intraductal secretioninto the periductal connective tissue [2, 5]. The non-puerperal retroareolar breast abscess is a low-grade infec-tion, caused by squamous metaplasia of ductal epitheliumand plugging of the ducts with keratin, leading tosecondary breast infection [35]. Granulomatous mastitis

is a relatively rare chronic inflammatory disease ofunknown etiology, which tends to occur in youngerwomen and also can clinically and radiographicallymimic invasive breast carcinoma [2, 36]. Acute breastinfection can range from mild focal inflammation todistinct abscesses [1, 2, 35].

Due to the limited published data about the character-istics of different mastitis types in MRM, we decided toinclude only acute breast infections in our study for a moreunique comparison with IBC. Nevertheless, there areseveral limitations of the study. One limitation of ourinvestigation is its retrospective design. Another limitationof our study is the exclusion of patients. Our intention wasto evaluate various kinetic and morphological MRMparameters and to obtain reliable imaging findings. There-fore, our study concept avoided all factors with thepotential to influence dynamic or morphological criteria.For this reason, patients having received chemotherapy,breast conservation, mastectomy or radiation therapy priorto MRM were excluded. Hormone effects during thesecond half of the menstrual cycle or caused by hormonereplacement therapy can affect the specificity of MRM [37,38]. Especially hormone replacement therapy is able tocomplicate the detection of proliferative or inflammatorynon-mass-like enhancements and of carcinomas in situ[38]. Therefore, patients were only enrolled in thisinvestigation if they had not received hormone replacementtherapy within the past 6 weeks. Premenopausal patientswere only considered if the MRM examination wasperformed in the first half of their menstrual cycle.

Another limitation of our investigation is a possibleselection bias with evaluation of patients suffering fromrelatively severe acute mastitis. It might be expected thatwomen with minimal inflammatory symptoms who aretreated successfully by their physicians are underrepre-sented. Otherwise, the patients in this study are representa-tive of those subjects who present the greatest diagnostic

Fig. 4 The histopathologicalimages (hematoxylin-eosin; ori-ginal magnification ×200) revealcharacteristics of an acute mas-titis (a) and an inflammatorycarcinoma (b; invasive ductal,G3). In the image of the mastitis(a), an infiltration of the mam-mary gland by macrophages andneutrophil granulocytes (arrows)are visualised. The image of theinflammatory carcinoma (b)reveals tumour emboli inlymphatic capillaries (circles)and infiltration of the stroma bytumour cells (lines)

2378

challenge because of similar clinical, X-ray mammogra-phical and sonographic features compared with inflamma-tory breast carcinoma.

The differential diagnosis of AM versus IBC remainsdifficult. Considering only limited MRM parameters, suchas the BI-RADS criteria, will not enable the definitediscrimination between the two diseases. However, con-sidering multiple dynamic and morphologcial criteriaseems to have the potential for a timely and precisedifferentiation. Because inflammatory carcinoma is a verymalignant form of invasive breast cancer with a pooroutcome, a histological diagnosis by needle or punch skinbiopsy is still necessary if any diagnostic doubt exists,especially if the clinical symptoms do not decrease duringantibiotic treatment within approximately one week [8, 9].Further investigations should prospectively analyse theclinical value of MRM in the differential diagnosis of

various lactational and non-lactational mastitis subtypescompared to IBC.

Conclusion

The combination of multiple dynamic and morphologicalMRM criteria seems to have the potential for an early andreliable discrimination between IBC and AM. However,the discrimination between the two diseases remains achallenge. Due to the serious consequences of a diagnosticdelay of IBC, a histological corroboration is necessary ifany doubt exists. In all patients with clinical inflammatorysigns, an initial MRM examination is valuable for apossible discrimination between IBC and AM as well as forthe interpretation of follow-ups by either confirming AMor, if biopsy could be false negative, by identifying IBC.

References

1. Crowe DJ, Helvie MA, Wilson TE(1995) Breast infection. Mammogra-phic and sonographic findings withclinical correlation. Invest Radiol30:582–587

2. Bässler R (1997) Mastitis. Classifica-tion, histopathology and clinicalaspects. Pathologe 18:27–36 DOI10.1007/s002920050193

3. Hance KW, Anderson WF, Devesa SS,Young HA, Levine PH (2005) Trendsin inflammatory breast carcinoma inci-dence and survival: the surveillance,epidemiology, and end results programat the National Cancer Institute. J NatlCancer Inst 97:966–975 DOI 10.1093/jnci/dji172

4. Günhan-Bilgen I, Üstün EE, Memis A(2002) Inflammatory breast carcinoma:mammographic, ultrasonographic,clinical, and pathologic findings in 142cases. Radiology 223:829–838 DOI10.1148/radiol.2233010198

5. Lequin MH, van Spengler J, van Pel R,van Eijck C, van Overhagen H (1995)Mammographic and sonographicspectrum of non-puerperal mastitis.Eur J Radiol 21:138–142 DOI10.1016/0720–048X(95)00699-Q

6. Iglesias A, Arias M, Santiago P,Rodríguez M, Mañas J, Saborido C(2007) Benign breast lesions thatsimulate malignancy: magneticresonance imaging with radiologic-pathologic correlation. Curr ProblDiagn Radiol 36:66–82 DOI 10.1067/j.cpradiol.2006.12.001

7. Lerebours F, Bieche I, Lidereau R(2005) Update on inflammatory breastcancer. Breast Cancer Res 7:52–58 DOI10.1186/bcr997

8. Rieber A, Tomczak RJ, Mergo PJ,Wenzel V, Zeitler H, Brambs HJ (1997)MRI of the breast in the differentialdiagnosis of mastitis versus inflamma-tory carcinoma and follow-up. JComput Assist Tomogr 21:128–132

9. Fischer U, Vosshenrich R, vonHeyden D, Knipper H, Oestmann JW,Grabbe E (1994) Inflammatory lesionsof the breast: indication for MR-mammography? Rofo 161:307–311

10. Orel SG, Schnall MD (2001) MRimaging of the breast for the detection,diagnosis, and staging of breast cancer.Radiology 220:13–30

11. Kuhl CK (2007) Current status ofbreast MR imaging. Part 2. Clinicalapplications. Radiology 244:672–691DOI 10.1148/radiol.2443051661

12. Kaiser WA, Zeitler E (1989) MRimaging of the breast: fast imagingsequences with and without Gd-DTPA.Preliminary observations. Radiology170:681–686

13. Heywang-Köbrunner SH, Viehweg P,Heinig A, Küchler C (1997) Contrast-enhanced MRI of the breast: accuracy,value, controversies, solutions. Eur JRadiol 24:94–108

14. Kuhl CK, Mielcareck P, Klaschik S,Leutner C, Wardelmann E, Gieseke J,Schild HH (1999) Dynamic breast MRimaging: are signal intensity timecourse data useful for differentialdiagnosis of enhancing lesions?Radiology 211:101–110

15. Sherif H, Mahfouz AE, Oellinger H,Hadijuana J, Blohmer JU, Taupitz M,Felix R, Hamm B (1997) Peripheralwashout sign on contrast-enhanced MRimages of the breast. Radiology205:209–213

16. Kuhl CK, Klaschik S, Mielcarek P,Gieseke J, Wardelmann E, Schild HH(1999) Do T2-weighted pulsesequences help with the differentialdiagnosis of enhancing lesionsin dynamic breast MRI? J MagnReson Imaging 9:187–196

17. Fischer DR, Wurdinger S, Böttcher J,Malich A, Kaiser WA (2005) Furthersigns in the evaluation of magneticresonance mammography: a retrospec-tive study. Invest Radiol 40:430–435

18. Malich A, Fischer DR, Wurdinger S,Böttcher J, Marx C, Facius M, KaiserWA (2005) Potential MRI interpreta-tion model: differentiation of benignfrom malignant breast masses. AJR AmJ Roentgenol 185:964–970 DOI10.2214/AJR.04.1073

19. Baum F, Fischer U, Vosshenrich R,Grabbe E (2002) Classification ofhypervascularized lesions in CE MRimaging of the breast. Eur Radiol12:1087–1092 DOI 10.1007/s00330–001–1213–1

20. Tomczak R, Rieber A, Zeitler H,Rilinger N, Kreienberg R, Brambs HJ(1996) The value of MR-mammogra-phy at 1.5 tesla in the differentialdiagnosis of non-puerperal mastitis andinflammatory breast carcinoma. Rofo165:148–151

2379

21. Belli P, Costantini M, Romani M,Pastore G (2002) Role of magneticresonance imaging in inflammatorycarcinoma of the breast. Rays 27:299–305

22. American College of Radiology (2003)Breast imaging reporting and datasystem (BI-RADS) atlas, 4th edn.American College of Radiology,Reston, VA

23. Fischer DR, Baltzer P, Malich A,Wurdinger S, Freesmeyer MG, Marx C,Kaiser WA (2004) Is the “bloomingsign” a promising additional tool todetermine malignancy in MRmammography? Eur Radiol 14:394–401 DOI 10.1007/s00330–003–2055–9

24. Renz DM, Baltzer PAT, Böttcher J,Facius M, Pfleiderer SOR, Gajda M,Camara O, Kaiser WA (2006) Newsigns of tumors in magnetic resonanceimaging. Eur Radiol 16(Suppl5):E80–E82 DOI 10.1007/s10406–006–0187–0

25. Friedman EP, Hall-Craggs MA,Mumtaz H, Schneidau A (1997) BreastMR and the appearance of the normaland abnormal nipple. Clin Radiol52:854–861 DOI 10.1016/S0009–9260(97)80081–5

26. Kazama T, Nakamura S, Doi O, SuzukiK, Hirose M, Ito H (2005) Prospectiveevaluation of pectoralis muscleinvasion of breast cancer by MRimaging. Breast Cancer 12:312–316DOI 10.2325/jbcs.12.312

27. Mahfouz AE, Sherif H, Saad A, TaupitzM, Filimonow S, Kivelitz D, Hamm B(2001) Gadolinium-enhanced MRangiography of the breast: is breastcancer associated with ipsilateral highervascularity? Eur Radiol 11:965–969DOI 10.1007/s003300000668

28. Buadu LD, Murakami J, Murayama S,Hashiguchi N, Sakai S, Masuda K,Toyoshima S, Kuroki S, Ohno S (1996)Breast lesions: correlation of contrastmedium enhancement patterns on MRimages with histopathologic findingsand tumor angiogenesis. Radiology200:639–649

29. Furman-Haran E, Schechtman E, KelczF, Kirshenbaum K, Degani H (2005)Magnetic resonance imaging revealsfunctional diversity of the vasculaturein benign and malignant breast lesions.Cancer 104:708–718 DOI 10.1002/cncr.21225

30. Sunderkötter C, Steinbrink K, GoebelerM, Bhardwaj R, Sorg C (1994)Macrophages and angiogenesis. JLeukoc Biol 55:410–422

31. Brown LF, Guidi AJ, Schnitt SJ, VanDe Water L, Iruela-Arispe ML, YeoTK, Tognazzi K, Dvorak HF (1999)Vascular stroma formation incarcinoma in situ, invasive carcinoma,and metastatic carcinoma of the breast.Clin Cancer Res 5:1041–1056

32. Taichman NS, Young S, Cruchley AT,Taylor P, Paleolog E (1997) Humanneutrophils secrete vascular endothelialgrowth factor. J Leukoc Biol 62:397–400

33. Macéa JR, Fregnani JHTG (2006)Anatomy of the thoracic wall, axillaand breast. Int J Morphol 24:691–704

34. Lee KW, Chung SY, Yang I, Kim HD,Shin SJ, Kim JE, Chung BW, Choi YA(2005) Inflammatory breast cancer:imaging findings. Clin Imaging 29:22–25

35. Haagensen CD (1986) Infections in thebreast. In: Haagensen CD (ed) Diseasesof the breast. 3rd edn. Saunders,Philadelphia, pp 384–393

36. Yilmaz E, Lebe B, Usal C, Balci P(2001) Mammographic and sono-graphic findings in the diagnosis ofidiopathic granulomatous mastitis. EurRadiol 11:2236–2240

37. Rieber A, Nüssle K, Merkle E,Kreienberg R, Tomczak R, Brambs HJ(1999) MR mammography: influenceof menstrual cycle on the dynamiccontrast enhancement of fibrocysticdisease. Eur Radiol 9:1107–1112

38. Pfleiderer SOR, Sachse S, Sauner D,Marx C, Malich A, Wurdinger S,Kaiser WA (2004) Changes in magneticresonance mammography due tohormone replacement therapy. BreastCancer Res 6:R232–R238 DOI10.1186/bcr779

2380