Clumped vs non-clumped internal enhancement patterns in linear non-mass enhancement on breast MRI

Abstract

Objective:

To compare positive predictive values (PPVs) of clumped vs non-clumped (homogenous and heterogeneous) internal enhancement on MRI detected linear non-mass enhancement (NME) on MRI-guided vacuum-assisted breast biopsy (MRI-VABB).

Methods:

With IRB (Institutional Review Board) approval, we retrospectively reviewed 598 lesions undergoing MRI-VABB from January 2015 to April 2018 that showed linear NME. We reviewed the electronic medical records for MRI-VABB pathology, any subsequent surgery and clinical follow-up. The X² test was performed for univariate analysis.

Results:

There were 120/598 (20%) linear NME MRI-VABB lesions with clumped (52/120, 43%) vs non-clumped (68/120, 57%) internal enhancement, average size 1.8 cm (range 0.6–7.6 cm). On MRI-VABB, cancer was identified in 22/120 (18%) lesions, ductal carcinoma in situ (DCIS) was found in 18/22 (82%) and invasive cancer in 4 (18%). 3/31 (10%) high-risk lesions upgraded to DCIS at surgery, for a total of 25/120 (21%) malignancies. Malignancy was found in 12/52 (23%) clumped lesions and in 13/68 (19%) of non-clumped lesions that showed heterogeneous (5/13, 38%) or homogenous (8/13, 62%) internal enhancement. The PPV of linear NME with clumped internal enhancement (23.1%) was not significantly different from the PPV of non-clumped linear NME (19.1%) (p = 0.597). The PPV of linear NME lesions <1 cm (33.3%) was not significantly different from the PPV of lesions ≥1 cm (18.6%) (p = 0.157).

Conclusions:

Linear NME showed malignancy in 21% of our series. Linear NME with clumped or non-clumped internal enhancement patterns, regardless of lesion size, might need to undergo MRI-VABB in appropriate populations.

Advances in knowledge:

Evaluation of linear NME lesions on breast MRI focuses especially on internal enhancement pattern.

Introduction

Clinical indications for contrast-enhanced MRI of the breast include screening of high-risk populations, evaluating extent of disease in patients with known breast cancer, or workup of abnormal imaging on mammography or ultrasound. ^1–3 Although breast MRI is the most sensitive imaging modality for detecting breast cancer, ⁴ it has lower specificity for breast cancer due to overlap of features of benign and malignant lesions. ^5,6 Because MR-guided biopsies are invasive, time-consuming and expensive, it is important to define criteria for suspicious breast masses needing biopsy.

The fifth edition of the Breast Imaging and Reporting Data System classifies abnormal enhancement into three types: focus, mass, and nonmass enhancement (NME). ⁷ Lesion type, abnormality morphology, and kinetics were used to discriminate benign from malignant lesions. ⁸ Linear distribution is of the BI-RADS distribution descriptors of NME and can be subclassified into internal enhancement patterns of clumped, homogeneous, heterogeneous, and clustered ring. ⁷ Among NME lesions, segmental and clumped linear enhancement patterns are seen more frequently as suspicious characteristics for malignancy. ⁹ However, the positive predictive values (PPVs) of linear NME varied among the previous studies, and the distribution and internal characteristics are often interwoven. ^9–12

Currently, there are few studies on linear NME lesions focusing only on internal enhancement pattern and research is required to identify the associations with malignancy on this topic. We aim to compare the PPVs of clumped internal enhancement characteristics versus non-clumped (homogeneous, heterogeneous, and clustered ring) on linear NME on MRI-guided vacuum-assisted breast biopsy (VABB).

Methods and materials

Study population

This Health Insurance Portability and Accountability Act (HIPAA) compliant single institution study was approved by our institutional review board and granted a waiver of informed consent. We retrospectively reviewed the imaging, pathology, and radiology reports of breast lesions in patients who underwent 9-gauge MRI-guided VABB at our institution from January 2015 to April 2018.

MRI acquisition

MRI was performed in the prone position with a 16-channel dedicated breast coil on a 3T system (GE Healthcare, Milwaukee, WI, USA). Each study included an axial T1W 3D gradient echo, axial T2W fat-suppressed images (CUBE), axial diffusion-weighted 2D EPI images with B = 0 and B = 600, and axial multiphase centrically encoded 3D T1W SPGR (Spoiled gradient recalled echo) images with Dixon separation of fat and water into separate images: a pre-contrast high-resolution image, 14 view shared rapid dynamic contrast-enhanced (DCE) images (15 sec each) with bolus power injection of 0.1 mmol/kg of gadobutrol (Gadovist, Bayer Health Care, Berlin, Germany) and a 20-ml saline flush 15 s after start of the acquisition, four high-resolution post-contrast phases (150 s each). Multiplanar reformation, thin slab maximal intensity projection (MIP) images and rotating MIP images were created from the first high-resolution post-contrast “peak” phase. Kinetic enhancement curves were performed by placing a region of interest on suspicious enhancing lesions using Aegis (Sentinelle Medical Inc., Toronto, Canada) or by DynaCAD (Invivo, Gainesville, FL, USA) system.

MRI-guided vacuum-assisted breast biopsies (VABB) were obtained on either a 1.5T or 3T scanner (Signa, GE Healthcare, Milwaukee, WI, USA) with 16-channel dedicated breast table top Sentinelle breast coils (Invivo, Gainesville, FL, USA) and open grid. Targeting was obtained using computer software (DynaCAD, Invivo, Gainesville, FL, USA, or Aegis, Sentinelle Medical, Inc, Toronto, Canada). Contrast enhancement of the target was confirmed after a bolus i.v. injection 0.1 mmol/kg of gadolinium contrast. All biopsies were performed with a 9-gauge VABB system (ATEC; Hologic Inc., Bedford, MA, USA), followed by placement of a titanium or MRI compatible stainless steel marker at the site of biopsy. For all patients, a post-biopsy mammogram was performed to confirm the position of the clip relative to the biopsy site.

MRI interpretation and data acquisition

All breast MRI studies were reported according to the fifth edition of the ACR BI-RADS MRI lexicon ⁷ and were initially interpreted by one of the seven fellowship trained MQSA-certified radiologists in the breast imaging department. Two breast-imaging radiologists (JC and SO) reviewed the MR images on a GE workstation to confirm that lesions undergoing MRI-guided VABB were comprised of linear non-mass enhancement, and noting that only lesions that were given a BI-RADS category four or five were biopsied. Lesion and patient characteristics were recorded in a HIPAA compliant data base including: examination indication, patient age, personal history of breast cancer, lesion size, internal enhancement characteristics, kinetic features prompting biopsy, MRI-guided breast VABB core biopsy results, and subsequent surgery results or >1 year imaging follow-up after biopsy. Lesions were separated into clumped and non-clumped (which include heterogeneous, homogeneous, clustered ring) internal enhancement characteristic categories. Enhancement kinetic curve descriptions were based on BI-RADS terminology of initial phase (slow, medium, and fast) and delayed phase (persistent, plateau, and washout), and recorded the most suspicious kinetic curve (initial, delay, or both), and recorded as described on the initial report by the original interpreting radiologist.

Statistical analysis

Statistical Package for Social Sciences (SPSS) for Windows, Statistics v.16.0. (SPSS Inc., Chicago, USA) was used to perform statistical analyses. The PPV of each internal enhancement characteristic was first calculated using standard averages and percentages. Exact 95% CI were calculated for each PPV. The X² test was used for univariate analysis. P-values less than 0.05 were considered statistically significant. Inter-observer agreement was examined using the intraclass correlation coefficient (ICC).

Results

Study populations

There were 598 MRI-VABB biopsies in the study period of which 120/598 (20%) showed linear NME, which comprise the study group. The ICC showed a good inter-observer reliability for the interpretation of linear NME on breast MRI (average measure of the ICC = 0.85). The indications of the breast MRI were screening of high-risk populations 60/120 (50%), staging 37/120 (30.8%), workup of abnormal imaging on mammography or ultrasound 22/120 (18.3%), and treatment-response assessment after neoadjuvant chemotherapy 1/120 (0.9%). Of these 120 linear NME lesions, there were 25 (20.8%) cancers, four invasive ductal carcinoma (IDC) and 21 ductal carcinoma in situ (DCIS) (Figure 1) (Table 1), and which will be described in detail subsequently. The mean patient age was 54.1 years old (ranged 27–82 years). The average number of core biopsies per lesion was nine cores (range: 4–17 cores). 53/120 (44%) NME lesions were operated upon. Of the remaining 67/120 (54%) NME not operated upon, two benign findings found at VABB were lost to follow up, and 65 (64 benign, one high-risk findings at VABB) showed no suspicious imaging and/or clinical features on follow up. The mean interval between VABB and imaging follow-up is 1.7 years.

Figure 1.

The pathological result of the 120 linear non-mass enhancement (NME) lesions undergoing vacuum-assisted breast biopsy (VABB). ALH, Atypical lobular hyperplasia; ADH, Atypical ductal hyperplasia; DCIS, Ductal carcinoma in situ; IDC, Invasive ductal carcinoma; LCIS, Lobular carcinoma in situ; RS/CSL, Radial scar/complex sclerosing lesion.

Table 1.

Internal Enhancement Patterns and Kinetic Curve in 120 findings of Linear Nonmass Enhancement on Contrast-Enhanced Breast MRI

Descriptor	Lesions N (%)	No. of cancer	PPV	p value	Histological diagnosis
					Invasive	DCIS
All linear NME	120 (100)	25	20.8	NA	4/25 (16.0)	21/25 (84.0)
Internal enhancement
Clumped	52 (43.3)	12	23.1	0.522	1	11
Homogeneous	33 (27.5)	8	24.2		3	5
Heterogeneous	35 (29.2)	5	14.3		0	5
Kinetic curve
Early a	67 (100)	10		0.243	1/10 (10)	9/10 (90)
Fast	54 (80.6)	9	16.7		0	9
Medium	6 (9.0)	1	16.7		1	0
Slow	7 (10.4)	0	0		0	0
Delayed b	85 (100)	16		0.754	2/16 (12.5)	14/16 (87.5)
Washout	33 (38.8)	7	21.2		0	7
Plateau	24 (28.2)	4	16.7		2	2
Persistent	28 (33.0)	5	17.9		0	5

DCIS, Ductal carcinoma in situ; NA, Not applicable.

^aUnknown Early Kinetic Phase n = 53.

^bUnknown Delayed Kinetic Phase n = 35.

Lesion characteristics on MR imaging

The average size of the linear NME lesions was 1.8 cm (range 0.6–7.6 cm). Of these, 52/120 (43%) had clumped internal enhancement characteristics and 68/120 (57%) were non-clumped. Non-clumped internal characteristics were heterogeneous, homogenous, and clustered ring, but there were no lesions with clustered ring internal enhancement in this population. Of the 68 lesions with non-clumped internal enhancement characteristics, 35/68 (51%) showed heterogeneous internal enhancement (Figure 2) and 33/68 (49%) showed homogeneous internal enhancement (Figure 3).

Figure 2.

47-year-old female underwent staging MRI, axial T ₁-weighted contrast-enhanced image showed a 1.4 cm non-mass enhancement (NME) linear distribution with non-clumped (heterogeneous) internal enhancement (arrow). MRI-guided vacuum-assisted breast biopsy showed lobular carcinoma in situ and atypical lobular hyperplasia.

Figure 3.

46-year-old female underwent staging MRI for right invasive ductal carcinoma. Sagittal maximum intensity projection showed a 1.7 cm non-mass enhancement (NME) linear distribution with non-clumped (homogeneous) internal enhancement (arrow). The pathology result was intermediate grade ductal carcinoma in situ.

Initial and delayed kinetic data were reported in 67 and 85 lesions, respectively. The initial phase was fast in 54/67 (81%) lesions, medium in 6/67 (9%) lesions, and slow in 7/67 (10%) lesion. The delayed phased showing washout 33/85 (39%), plateau 24/85 (28%), and persistent 28/85 (33%) kinetics (Table 1).

On MRI-VABB, there were 67/120 (56%) benign lesions, 22/120 (18%) malignancies (18 DCIS, 3 IDC, one invasive lobular carcinoma), 31/120 (21%) were high-risk lesions. Of the high-risk lesions, 3/31 (17%) upgraded to DCIS at surgery (Figure 4), for a total of 25/120 (21%) malignancies. Malignancy was found in 12/52 (23%) lesions with clumped internal enhancement and in 13/68 (19%) of lesions with non-clumped internal enhancement characteristics. Of the linear non-clumped lesions with malignancy, 5/13 (38%) had heterogeneous internal enhancement, and 8/13 (62%) had homogeneous internal enhancement. Kinetic data of the 25 patients with biopsy proven malignancy were evaluated and initial and delayed kinetic data were available in 10 and 16 lesions. The largest number of these malignant lesions showed fast initial Phase 9/10 (90%) followed by washout delayed Phase 7/16 (44%) (Table 1).

Figure 4.

79-year-old female underwent screening MRI, sagittal T ₁-weighted contrast-enhanced image (A) showed a 2.6 cm non-mass enhancement (NME) linear distribution with clumped internal enhancement (arrow). Maximum intensity projection (MIP) of post MRI-guided vacuum-assisted breast biopsy (B). Susceptibility artefact from the biopsy clip is present (arrowhead). The pathology result was atypical ductal hyperplasia and excision yielded to intermediate grade ductal carcinoma in situ.

Positive predict value of linear NME

Overall, the PPV of NME with linear distribution was 20.8% (25/120; 95% CI, 13.5–28.1). The PPV of linear NME with clumped internal enhancement (23.1%; 95% CI, 11.6–34.6) was not significantly different from the PPV of non-clumped linear NME (19.1%; 95% CI, 9.8–28.4) (p = 0.597). The PPV of linear NME lesions <1 cm (33.3%; 95% CI, 11.5–55.1) was not significantly different from the PPV of lesions ≥1 cm (18.6%; 95% CI, 11.1–26.2) (p = 0.157). There was no association with malignancy between clumped and non-clumped NME in combination with size or suspicious kinetic data (Table 2).

Table 2.

PPVs for Malignancy in Linear Distributed NME on Contrast-enhanced MRI according to lesion characteristics and kinetic curves

Descriptors	Benign lesions (n)	Malignant lesions (n)	PPV (%) a	p value
Clumped	40	12	23.1 (11.6, 34.6)	0.597
Non-clumped	55	13	19.1 (9.8, 28.4)
NME <1 cm	12	6	33.3 (11.5, 55.1)	0.157
NME ≥ 1 cm	83	19	18.6 (11.1, 26.2)
Clumped linear <1 cm	6	2	25.0 (0.0, 55.0)	0.544
Clumped linear ≥1 cm	35	9	20.5 (8.6, 32.4)
Non-clumped linear <1 cm	7	3	30.0 (1.6, 58.4)	0.336
Non-clumped linear ≥1 cm	47	11	19.0 (8.9, 29.1)
Clumped and fast	17	2	10.5 (0.0, 24.3)	0.412
Non-clumped and fast	29	6	17.1 (4.6, 29.6)
Clumped and washout	10	2	16.7 (0.0, 37.8)	0.494
Non-clumped and washout	16	5	23.8 (5.6,42.0)

^aNumbers in parentheses are 95% confidence intervals.

Discussion

The fifth edition of the ACR BI-RADS lexicon made specific changes to the distribution and internal enhancement patterns of NME. ¹³ With regard to distribution, the word “ductal” was replaced with “linear” to describe enhancement arrayed in a line or a line that branches. ⁷ Our study showed that linear NME on breast MRI may represent malignancies such as IDC, invasive lobular carcinoma, DCIS or various benign breast processes as previous studies. ^14–18 The PPVs of linearly distributed NME show range from 11 to 67% in the literature. ^{9,10,12,14,19} In our study of BIRADS four or five linear NME lesions leading to MRI-guided VABB, we found that linear NME had a PPV of 20.8%. As a result of this variability, studies of additional features of linear NME were performed to determine if specificity could be improved. Machida et al studied the PPVs of branching verses non-branching linear NME, showing that a branching pattern was a significantly stronger predictor of malignancy than was the linear pattern. ²⁰ Combining these features may be used to improve the diagnosis of linear NME lesions.

To our knowledge, this is the only study that looked specifically at the PPVs of the internal enhancement characteristics of linear NME. Our results show no significant difference in PPVs of clumped linear non-mass enhancement (23.1%) vs non-clumped (19.1%) – heterogeneous (14.3%) or homogenous (24.2%) –- internal enhancement characteristics. In review of previous studies, most of currently published articles evaluating the PPVs of malignancy for linear distribution and associated internal enhancement were performed before the publication of the updated lexicon (Table 3). The clumped enhancement had been mixed with ductal or branching pattern, which may explain the wide range of PPV of overall clumped lesions in these studies, ranging from 20 to 88%. A few studies have performed a detailed analysis of linear clumped enhancement. Liberman et al reported that the PPV of linear clumped lesion ranged 31–35%. ^10,12 A higher PPV of clumped enhancement was demonstrated by Morakkabati et al (PPV = 59%); however, in that study, NME included segmental and linear distributions and may have had an inherent bias toward malignancy. Tozaki and Fukuda ⁹ found extremely high PPV for linear clumped NME, with malignancy found in 3 of 4 (75%) lesions; however, the study had small numbers of both total biopsied NME and clumped lesions. Our results reflect the results of the highest number of total biopsied linear distribution (n = 120) and linear clumped lesions (n = 52) of all studies published to date (Table 3).

Table 3.

Review of Literature on Malignancy vs Benign Pathology in Linear Non-Mass Enhancement (NME) and Stratified by Internal Enhancement Characteristics

Study		Total No. of Biopsied NMEs	No. of linear/Total No. of NMEs (%)	No. of Malignant Linear NMEs/Total linear NME(%)	No. of Malignant clumped/Total clumped NME (%)	No. of Malignant linear clumped/Total linear clumped (%)
Lead Author	Year
Liberman et al 10	2002	40	21/40 (53)	5/21 (24)	9/22 (41)	5/16 (31)
Liberman et al 12	2003	150	88/150 (59)	23/88 (26)	NA	18/52 (35)
Morakkabati et al 11	2005	38	NA	2/10 (20)	NA	10/17 (59) a
Tozaki et al 9	2006	30	8/30 (27)	2/6 (33)	7/8 (88)	3/4 (75)
Sakamoto et al 19	2008	102	9/102 (9) b	1/9 (11)	2/10 (20)	NA
Uematsu et al 14	2012	124	9/124 (7)	6/9 (67)	26/32 (81)	NA
Ballesio et al 15	2014	94	19/94 (20)	12/19 (63)	9/14 (64)	NA
Chikarmane et al 21	2017	144	38/144 (26)	12/38 (32)	11/33 (33)	NA
Our Study	2019	598	120/598 (20)	25/120 (21)	NA	12/52 (23)

NA, Not available.

^aIncluding both linear and segmental distribution.

^bAssessed by Linear-ductal pattern.

We found that the PPV of linear NME <1 cm (33.3%) was not significantly different the PPV of lesions ≥ 1 cm (18.6%), comparable to the result by Gutierrez et al that size was not a significant predictor of malignancy for NME. ²² This finding counters the conclusions of Machida et al who concluded that NME lesions with a linear pattern that are smaller than 1 cm can be managed with follow-up. ²⁰ One possible reason for this difference is that our population was different than that of the Machida et al in that we only studied patients who underwent MRI-guided biopsy and all patients were at very high risk for breast cancer. This suggests, in the appropriate population (as in 80% of our population are either high-risk screening or staging for known breast cancer), linear NME might need to be biopsied rather than observed even if the lesion size is less than 1 cm.

Studies have shown that most common malignant causes of NME are DCIS and diffuse invasive breast cancers, particularly lobular cancer, but also occasionally ductal cancers. ^23,24 In particular, DCIS most commonly manifests as NME, less frequently as a mass or a focus. ²⁵ Rosen et al showed that pure DCIS lesions show NME in 59% of cases, whereas 14% present as an enhancing mass, 14% show no enhancement, and 12% as a focus. ²⁶ It has been suggested that DCIS appears as NME because DCIS typically follows the ductal system, which means the most frequent enhancement was within a segmental or linear (ductal) distribution, and internal enhancement is usually clumped. ^10,11,27,28 Our results supported these findings demonstrated in previous studies in that 11/21 (52%) of our linear NME DCIS showed clumped internal enhancement, higher than either heterogeneous (5/21, 24%) or homogeneous (5/21, 24%) internal enhancement if each of the non-clumped patterns were compared alone. Overall, our results showed most malignant linear distributed NME were DCIS (21/25, 84%), and many fewer were invasive cancer (3/25 were IDC, and only one invasive lobular carcinoma).

The classically suspicious kinetic curve for a lesion detected at MR imaging includes fast early enhancement and/or delayed washout, or plateau-type enhancement. ^29,30 Kinetic information has proved to be helpful in the differential diagnosis of MR imaging – detected mass lesions. ²⁹ However, the variability of kinetic curves in both benign and malignant NME has led a number of investigators to consider kinetic features unreliable for a diagnosis of malignancy, particularly for DCIS. ^31–33 This is because kinetic curve is influenced by several pathological factors, including the extend and pattern of vascularization, vessel permeability, cellularity, interstitial pressure, and the fraction of the extracellular space. ³² Jansen et al evaluated 852 MRI detected lesions (552 mass, 261 NME, and 39 focus) and reported that in NME, only washout parameters may be relevant. ²⁴ In our study, fast-washout pattern was the most frequent descriptor for malignant pathology in linear NME, and most are DCIS. In DCIS, a fast-washout pattern may associate with a high density of ducts, an abundance of blood vessels, and a high degree of inflammatory cell infiltration. ³³ However, none of the kinetic features were significant predictors of carcinoma when compared to the entire study population, which was similar to Liberman et al ¹² study

Our study has some limitations. First, it is a retrospective study and we examined a high-risk patient population and included only lesions referred for biopsy, which may lead to selection bias. Second, there was no clustered ring enhancement (CRE) in our study. CRE can be difficult to differentiate from other internal enhancement patterns and may be present with either clumped or heterogeneous NME. This challenge raises the question of whether CRE should be assessed differently, such as assessed as categorically present or not present, to reflect clinical practice, and possibly be listed as an associated feature in the BI-RADS atlas. ²¹

Conclusions

Linear NME showed malignancy in 21% of our series. Our recommendations would be, in specific patient groups, such as screening in females at high-risk for breast cancer or staging for known breast cancer, linear NME with clumped or non-clumped internal enhancement patterns might need to undergo MRI-VABB regardless of lesion size or type of kinetic curve.