Assessment of Pectoralis muscle invasion using combined DBT and contrast-enhanced mammography: Retrospective multi-reader study

Abstract

Objectives:

To evaluate combined digital breast tomosynthesis and contrast-enhanced mammography (DBT/CEM) for predicting pectoralis muscle invasion.

Methods:

This retrospective multi-reader cohort study included research patients who underwent combined DBT/CEM for breast cancer staging and had prepectoral masses. Images were independently reviewed by six fellowship-trained breast radiologists. Diagnostic performance, reader confidence, and inter-reader agreement were calculated for each image type/modality.

Results:

Among 10 patients with prepectoral masses on DBT/CEM, muscle invasion was present in 3 and absent in 7. The overall diagnostic accuracy of DBT/CEM for PMI was 0.6 (range 0.4–0.9); for predefined radiologic signs it was 0.5–0.7 for low energy (LE) CEM, 0.4–0.7 for DBT, and 0.4–0.8 for recombined (RC) CEM. Muscle deformity on MLO views had the highest accuracy (0.7–0.8). On a scale of 1–3, mean radiologist confidence for combined DBT/CEM was 1.9 (1.5–2.3; SD=0.65). Median confidence ranged from 1.9 for RC to 2.2 for DBT. Per-case reader agreement was poor (K=−0.01) for DBT/CEM; poor to slight (K= −0.13–0.40, median 0.28) for RC; slight to fair (K = 0.04–0.43, median 0.27 and K = 0.02–0.42, median 0.19, respectively) for DBT and LE. In two patients with subpectoral breast implants CEM was accurate in PMI detection, while MRI had one false-positive result.

Conclusion:

Combined DBT/CEM accuracy and inter-reader agreement are suboptimal for PMI evaluation, except in patients with breast implants. RC images marginally improve accuracy compared to LE images but have lowest radiologist confidence. DBT has lowest accuracy but highest confidence. Muscle deformity on MLO view was the most accurate sign.

Critical Relevance Statement:

Combined DBT/CEM demonstrated suboptimal diagnostic accuracy, reader confidence, and inter-reader agreement for detecting pectoralis muscle invasion (PMI) in prepectoral breast cancer (BC) except for patients with subpectoral breast implants, where recombined images on implant-displaced CEM views performed better than MRI.

Introduction

Clinical significance

The clinical significance of isolated pectoralis muscle invasion (PMI) by posteriorly located breast cancer (BC) and chest wall invasion differ. Direct chest wall invasion, defined as a tumor infiltrating the ribs, intercostal muscles, and/or serratus anterior muscle, is classified as the minimum T4a tumor stage, IIIB disease stage and carries a worse prognosis compared to a cancer without chest wall invasion.^1–3 In contrast, isolated PMI does not affect BC staging or prognosis, but impacts surgical planning by facilitating preoperative estimation of tissue resection required to achieve negative histologic margins. Surgical margin status directly impacts the rate of in-breast and locoregional recurrence in the muscle.^4–7

Conventional imaging methods

Full field digital mammography (FFDM) is the first line of imaging in BC staging, and the majority of BC patients receive a mammogram at initial presentation.⁸ Since the pectoralis muscle is frequently only partially visible on FFDM, FFDM is suboptimal for the estimation of PMI. Posterior breast cancers inseparable from the underlying pectoralis muscle and posterior trabecular thickening may be subtle clues to possible PMI.^3,9 Breast ultrasound (US) is equally unreliable for diagnosing isolated PM invasion in clinical practice. Some limitations of US for this application include posterior acoustic shadowing from a mass, similar echogenicity of a cancer and the underlying muscle, inadequate ultrasound technique, and patient positioning (in a supine patient a posteriorly located cancer can abut and indent the muscle even in the absence of invasion).^3,10

Breast MRI is currently considered the standard imaging modality for predicting pectoralis muscle invasion,^3,11,12 prompting some experts to recommend breast MRI for chest wall evaluation in eligible patients.^1,13 The most specific MRI sign of pectoralis invasion is direct extension of an enhancing tumor into the underlying muscle with resulting muscle enhancement^{3,11, 12}; however, the absence of muscle enhancement does not exclude invasion.¹⁴ Additional MRI signs potentially contributing to a diagnosis of pectoralis muscle invasion but carrying variable diagnostic values are diffusion restriction in the muscle, muscle deformation, loss of prepectoral fat, and enhancement of the pectoralis fascia.^11,12,14

Contrast-enhanced mammography

Contrast-enhanced mammography (CEM) is a functional 2-dimensional mammographic technique that improves the diagnostic accuracy of digital mammography and is comparable in sensitivity, but superior in specificity, to breast MRI.^15–17 Contrary to MRI, CEM is a 2D technology and is susceptible to tissue superimposition that can affect both low energy (LE) and recombined (RC) images.¹⁸ LE images of CEM are comparable to full-field digital mammography (FFDM) and are interpreted as such.^19,20 FFDM (and, by analogy, LE) images are suboptimal for evaluation of pectoralis muscle invasion by posteriorly located breast cancers.⁹ Digital breast tomosynthesis (DBT) is a pseudo-3D imaging modality which increases tissue separation in the Z-axis (depth), improving lesion localization and margin characterization.^21,22

Purpose

We aimed to investigate whether CEM and DBT performed under the same compression would result in a superior prediction of pectoralis invasion by breast cancer compared to LE images alone. We hypothesized that a combination of contrast enhancement information on CEM with the improved lesion localization and margin characterization of DBT could lead to better diagnostic accuracy.

Methods

Patient population

This institutional HIPAA-compliant IRB-approved multi-reader study retrospectively analyzed prospectively collected data on consecutive patients with untreated breast cancer who underwent staging combined DBT/CEM under IRB-approved research protocols at a single tertiary referral center from 4/1/2021 to 4/1/2023. A waiver of informed consent was granted by the IRB. The presence of a prepectoral breast mass on CEM was prospectively recorded in a research database. A prepectoral mass on CEM was defined as a mass overlying or immediately adjacent to the pectoralis muscle with no clear intervening fat plane on any view.

Ground truth

The gold standard was the presence or the absence of PMI on surgical pathology (SP) in patients who had no neoadjuvant chemotherapy (NAC), or the presence of PMI on SP post NAC. For patients who underwent pre-treatment breast MRI and had NAC with therapy response and no definite PMI on SP, radiologists’ consensus on the presence or absence of muscle enhancement on pre-NAC MRI was used as a surrogate for the ground truth. Muscle enhancement was chosen as it was proven to represent the most specific sign of pectoralis invasion in previous studies.^11,12 Patients with no verifiable ground truth (lost to follow-up, surgery performed at outside facilities, no pre-treatment MRI in patients’s status post NAC) were not eligible for the study.

CEM imaging technique

Following institutional contrast eligibility screening procedures, an IV catheter was placed by a nurse in patient’s antecubital or forearm vein in a mammography suite immediately before the procedure. Low-osmolality iodine-based intravenous contrast (Omnipaque 350 mg I/mL; GE Healthcare) was injected using a power injector, at a rate of 2–3 mL/s for a total dose of 1.5 mL/kg body weight, but not exceeding 150 mL, followed by a 30 mL saline flush.

Imaging was done using an FDA-approved FFDM/DBT system capable of obtaining low- and high-energy exposures of the breast and producing subtracted contrast images (Senographe Pristina with Seno-Bright HD Gen II software [GE Healthcare, Buc, FR] or Selenia Dimensions with I-View 2.0 software [Hologic, Marlboro, MA]). The equipment could obtain DBT and CEM images under the same compression.

Imaging began 2 min after the end of the contrast injection and consisted of DBT projections followed by high and low energy 2D exposures in succession under the same compression. Both breasts were imaged in standard craniocaudal (CC) and mediolateral oblique (MLO) projections, starting with the breast of interest, alternating the breasts.

The breast of interest was also imaged with DBT and CEM in the lateromedial (LM) projection (Fig 1). The protocol was modified for the 2 patients with implants as follows: both breasts were initially imaged with FFDM in the MLO and the CC implant-in-place projections, followed by a contrast injection and the acquisition of alternating CC and MLO implant-displaced projections under the same compression, starting with the breast of interest, and including an implant-displaced LM view of the breast of concern.

Fig. 1.

Imaging sequence example for left breast cancer staging.

The images were immediately evaluated by the technologist for quality assurance. Additional full field views (technical repeat or exaggerated craniocaudal) were occasionally obtained by the technologists to compensate for motion or incomplete tissue coverage. Targeted additional images (spot compression DBT or CEM or spot compression magnification) were ordered by the radiologists, if necessary.

The patients were observed for 30 min after the contrast injection to ensure the absence of symptoms of a delayed contrast reaction. The patients were encouraged to drink at least 250 ml of water before and after the procedure.

MRI Imaging technique

Breast MRI examinations were performed on one of three systems: 1.5T SIGNA Artist (GE Healthcare, Chicago, IL, USA), 3T DISCOVERY 750 (GE Healthcare), or 3T Vida (Siemens Healthineers, Erlangen, Germany). The data was acquired with a 16- or an 18-channel dedicated bilateral breast coil. While there was some heterogeneity in the imaging protocols, the protocols generally included axial pre-contrast and sagittal post-contrast T1-weighted 3D sequences (0.56 – 0.99 mm³ voxel size). Axial T1-weighted 3D sequences (0.67 – 1.18 mm3 voxel size, fat saturated, temporal resolution < 2 min/phase) were acquired before and after the intravenous injection of 0.1 mmol/kg Gadobutrol (Bayer HealthCare, Leverkusen, Germany) at a rate of 2 mL/s, followed by a 30 mL saline flush. Early and late phase subtraction and maximum-intensity-projection images were generated from the dynamic images. Axial T2-weighted 2D images (3–4 mm slice thickness, 0.50 – 1.0 mm2 pixel area, fat saturated) and diffusion-weighted images (5 mm slice thickness, 4 – 4.5 mm2 pixel area, b-values of 100 and 800) were acquired after the injection of intravenous contrast material.

Pathology

Clinical imaging-guided needle biopsy and surgical pathology reports were reviewed using electronic heath records. Breast cancer type, the presence of PMI, the degree of neoadjuvant therapy response of the prepectoral cancer, and the presence of post-treatment changes involving the pectoralis muscles were extracted from the reports. In addition, surgical pathology specimens were evaluated by a fellowship-trained breast pathologist (AC) with a focus on PMI.

Reader study

The images were independently reviewed by six subspecialty-trained breast radiologists with 6–34 years of experience (median 13). All radiologists underwent CEM training with an educational module consisting of didactic material and 43 cases. In addition, before the study onset the radiologists had an independent CEM interpretation experience range of 8–205 cases (median 29). The readers independently sequentially reviewed CEM (LE and RC) and DBT images of the involved breast in the CC and the MLO projections with no reference to prior images. Using an Excel spreadsheet (Microsoft, Redmond, WA, USA), the readers evaluated and recorded several pre-defined signs of PMI and indicated their per-view and per-case estimation of the presence or absence of PMI for each component of the combined DBT/CEM study. Due to the absence of defined criteria of PMI on CEM, we adopted previously validated MRI signs of PMI^3,11,12,14 (Table 1). Subsequently, the radiologists reviewed the most recent breast MRI (within 30 days of CEM, before treatment) and recorded the presence or absence of pre-defined previously validated MRI imaging signs of PMI (Table 1). For each radiologic sign and each PMI estimate the radiologists provided their level of interpretative confidence (1=low, 2=intermediate, 3=high).

Table 1.

Parameters independently evaluated by 5 radiologists and used for statistical analysis.

CEM CC VIEW														OVERALL CC CEM MUSCLE INVASION		CEM MLO View														OVERALL MLO CEM MUSCLE INVASION		PER CASE RESULT CEM INVASION		MRI
CC FAT PLANE						CC Muscle Deformity						CC Muscle enhancement				MLO Fat Plane		MLO Muscle deformity						Muscle enhancement										Fat plane		Muscle deformity		Fascia enhancement		Muscle enhancement
On LE (yes∕no)	Confidence for LE	On RC (yes∕no)	Confidence for RC	On DBT (yes∕no)	Confidence for DBT	On LE (yes∕no)	Confidence For LE	On RC (yes/no)	Confidence for RC	On DBT (yes∕no)	Confidence for DBT	On RC (yes/no)	Confidence for RC	Y/N	Confidence	On LE (yes∕ no)	Confidence for LE	On RC (yes/no)	Confidence for RC	On DBT (yes/no)	Confidence for DBT	On LE (yes/no)	Confidence for LE	On RC (yes/no)	Confidence for RC	On DBT (yes/no)	Confidence for DBT	On RC (yes/no)	Confidence for RC	Yes/no	Confidence	Yes/no	Confidence	Yes/no	Confidence	Yes/no	Confidence	Yes/no	Confidence	Y/N	Confidence

Statistical analysis

Diagnostic performance, reader confidence, and inter-reader agreement for each image type/modality were calculated per radiologic sign, per imaging view, and per case. The median and range for these statistics were summarized across all readers. Inter-reader agreement was assessed using Fleiss’ kappa statistic. Forest plots were employed to illustrate the overall mean of reader confidence, along with its 95 % confidence interval, as well as the kappa statistics with their respective 95 % confidence intervals. The relationship between reader accuracy for PMI and both the number of CEM studies interpreted and years of experience was analyzed using Spearman’s rank correlation coefficient. All statistical analyses were performed using R (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria).

Results

Of 146 consecutive research patients with BC who underwent staging combined DBT/CEM, 10 had prepectoral cancers with verifiable ground truth for PMI. The patients were 39–75 years old (median 57). Two patients had intact subpectoral saline breast implants. Four patients underwent surgical treatment with no NAC and had no PMI on SP. Six patients underwent NAC. One post-NAC patient had partial response with residual PMI on SP. Five post-NAC patients had PCR with no PMI on SP, necessitating the use of breast MRI as a representation of the ground truth. The radiologists’ MRI consensus was the presence of PMI in 2 and the absence of PMI in 3 cases (Fig. 2).

Fig. 2.

Study cohort and distribution of positive (pectoralis muscle invasion (PMI) present) and negative (PMI absent) cases with the sources of the ground truth.

Two patients did not have MRI but underwent surgery without NAC and had no muscle invasion on SP. All patients had invasive cancer (9 invasive ductal cancer (IDC), 1 mixed invasive ductal/lobular cancer). Patient characteristics and outcomes are presented in Table 2.

Table 2.

Patient characteristics and outcomes.

N	Age	Breast Density	Ground Truth		Biopsy Pathology	Surgery Type	Surgical Pathology	Notes
			Invasion Yes (1)/No (2)	Source MRI (1)/SP (2)
1	75	B	1	1	IDC, grade 3	Mastectomy, total	Post NAC. Complete response, no chest wall invation
2	39	C	1	1	Mixed ductal and lobular, grade 1	Mastectomy, total	Post NAC. Residual multifocal Ca, no  chest wall invasion
3	40	C	1	1	IDC, grade 3	Segmental mastectomy	Post NAC. No residual tumor.
4	56	C	1	2	IDC with DCIS, grade 2	Mastectomy, total	No NAC. No muscle invasion.
5	77	B	2	1	IDC, grade 3	Mastectomy, total	Post NAC. No muscle invasion post response.
6	63	B	1	2	IDC with DCIS, grade 1	Segmental mastectomy	No NAC, no muscle invasion.
7	39	D	2	1	IDC with DCIS, grade 2	Mastectomy, total	Post NAC, no muscle invasion
8	62	C	1	2	IDC, grade1	Segmental mastectomy	No NAC, no muscle invasion.
9	59	B	1	2	IDC, grade 3	Mastectomy, total	No NAC, no muslce invasion.	Saline implants
10	48	C	2	2	IDC, grade 3	Mastectomy, total	Post NAC. Residual Ca with muscle  invasion.	Saline implants

NAC- neoadjuvant therapy. IDC- invasive ductal carcinoma. DCIS-ductal carcinoma in-situ. SP- surgical pathology, NAC- neo-adjuvant chemotherapy, Ca-cancer.

The overall median accuracy of combined DBT/CEM for PMI across all readers for all imaging signs and all image types was 0.6 (range 0.4–0.9). A sub-analysis of combined DBT/CEM by image type revealed a median accuracy of LE at 0.6 (range 0.5–0.7), a median accuracy of DBT at 0.5 (range 0.4–0.7), and a median accuracy of RC at 0.7 (range 0.4–0.8). The most accurate radiologic signs on RC were muscle deformity on the MLO view and muscle enhancement on the MLO view with accuracies 0.8 and 0.7, respectively. The single most accurate radiologic sign of PMI on both LE and DBT was muscle deformity on MLO view (accuracy of 0.7 on both modalities).

The overall median accuracy of MRI across all readers and all imaging signs was 0.8 (range 0.6–0.9). The most accurate MRI sign was muscle enhancement, followed by fascia enhancement (accuracies 0.8 and 0.7, respectively) (Table 3).

Table 3.

Ranges and medians of accuracy per image type and radiologic sign across all readers.

Modality  Radiologic Sign	CC Fat Plane	MLO Fat Plane	CC Muscle Deformity	LE  MLO Muscle Deformity	-	-
Accuracy Range	0.3–0.7	0.5–0.9	0.3–0.8	0.7–0.7	-	-
Median Accuracy	0.5	0.6	0.6	0.7	-	-
Modality				DBT
Radiologic Sign	CC Fat Plane	MLO Fat Plane	CC muscle deformity	MLO muscle deformity	-	-
Accuracy range	0.2–0.6	0.4–0.7	0.4–0.6	0.6–0.8	-	-
Median Accuracy	0.4	0.6	0.4	0.7	-	-
Modality				RC
Radiologic Sign	CC Fat Plane	MLO Fat Plane	CC Muscle Deformity	MLO Muscle Deformity	CC Muscle Enhancement	MLO Muscle Enhancement
Accuracy Range	0.3–0.4	0.5–0.6	0.6–0.8	0.7–0.8	0.7–0.8	0.6–0.8
Median Accuracy	0.4	0.5	0.7	0.8	0.7	0.7
				CONCLUSION FOR PMI
Views	CC View (LE+DBT+RC)		MLO View (LE+DBT+RC)		Total Per Case (both views, (LE+DBT+RC))
Accuracy Range	0.6–0.7		0.5–0.8		0.4–0.9
Median Accuracy	0.6		0.5		0.6
				MRI
Radiologic Sign	Absence of Fat Plane		Muscle Deformity	Fascia Enhancement		Muscle Enhancement
Accuracy Range	0.5–0.7		0.5–0.7	0.6–0.9		0.6–0.9
Median Accuracy	0.6		0.6	0.7		0.8

The source data is presented in supplemental materials (SEM Table 1). LE- low energy images of CEM, RC-recombined images of CEM, CC- craniocaudal projection, MLO- mediolateral oblique projection.

Reader confidence was the highest for MRI, ranging from 2.5 (fascia enhancement) to 2.8 (muscle deformity). The range of reader confidence for LE was 2.0–2.1 (median 2.0), for DBT 2.2–2.3 (median 2.2), and for RC 1.8–2.2 (median 1.9). The lowest confidence on RC was for muscle deformity on the MLO view (1.8), and the highest was for the presence of a fat plane on the CC view (2.2) (Fig. 3).

Fig. 3.

Ranges of confidence for the presence of muscle invasion across all readers per imaging sign, per modality, and per case. The vertical dashed line represents the overall average.

Inter-reader agreement for per-case PMI estimates on combined DBT/CEM studies was poor (kappa= −0.01, range −0.17–0.15). Per-view and per-imaging sign analysis demonstrated that the agreement for DBT and LE was slight to fair (K = 0.04–0.43, median 0.27 and K = 0.02–0.42, median 0.19, respectively). The agreement for RC was poor to slight (K= −0.13–0.4, median 0.28). Inter-reader agreement for MRI was fair to substantial (K = 0.40–0.61, median 0.6) (Fig. 4).

Fig. 4.

Ranges of agreement across all readers per imaging sign, per modality, and per case. Kappa values and corresponding agreement levels: <0 = poor; [0.01,0.20] = slight; [0.21,0.40] = fair; [0.41,0.60] = moderate; [0.61,0.80] = substantial; [0.81,1.00] = almost perfect. The vertical dashed line represents the overall average.

Two cases with subpectoral breast implants could not be statistically analyzed due to the small sample size. One patient went directly to surgery with no NAC and had no PMI on SP. The other patient was post-NAC but had residual cancer with PMI on SP. Combined DBT/CEM correctly diagnosed the absence of PMI in the first case and the presence of PMI in the second case. The presence or absence of muscle enhancement on the RC views and of muscle deformity on the RC MLO views were the most accurate signs. Reader confidence for the overall combined DBT/CEM assessment for PMI was 2.3 in the case with no PMI and 1.7 in the case with PMI. MRI demonstrated muscle enhancement with a high degree of confidence in both the false-positive and the true-positive cases (2.5 and 2.8, respectively) (Table 4).

Table 4.

Sub-analysis of 2 cases with subpectoral saline implants. The table contains radiologists’ consensus on the presence of a radiologic sign (2=sign is present; 1 =sign is absent) and the level of interpretative confidence (1 =low, 2=intermediate; 3=high) for each imaging modality and each projection. No consensus (50/50 split) was achieved by radiologists for the following signs: the presence of fat plane on the MLO DBT view for case 1; the presence of a fat plane on the CC DBT view, and muscle deformity on the MLO DBT view for case 2 (consensus 1.5).

Study		DBT/CEM combination																		MRI
View		CC View								MLO view									PMI Per Case
Imaging Sign		Fat Plane			Muscle Deformity			Muscle Enhancement	Per-view PMI	Fat Plane			Muscle Deformity			Muscle Enhancement	Per-View PMI			Fat Plane	Muscle Deformity	Fascia Enhancement	Muscle Enhancement
Modality		LE	RC	DBT	LE	RC	DBT	RC	LE+RC+DBT	LE	RC	DBT	LE	RC	DBT	RC	LE+RC+DBT
Casel (No PMI)	Consensus	2	2	2	1	1	1	1	1	2	2	1.5	1	1	1	1	1	1		1	1	2	2
	Average confidence	2.7	2.5	2.7	2.7	1.8	2.2	2.2	2.5	2.5	2.2	2.5	2.3	1.8	2.3	2.2	2	2.3		2.8	2.3	2.7	2.5
Case 2  (PMI)	Consensus	2	1	1.5	1	1	1	2	1	2	1	1	1	2	1.5	2	2	2		1	2	2	2
	Average confidence	1.8	2.5	2.2	2.5	1.8	2.3	1.5	1.5	1.7	1.8	2.3	1.7	1.8	2	1.5	1.7	1.7		2.8	2.3	2.8	2.8

We analyzed the diagnostic performance and the interpretative confidence of all participating radiologists based on their experience in breast imaging (range 6–34 years) and the degree of experience with CEM based on the number of studies read (range 8–205 cases). There was no evidence of improvement of the diagnostic accuracy of radiologists as a function of increasing experience with CEM or with breast imaging in general (SEM, Figs. 1,2).

Discussion

Overall, the combination of DBT and CEM provides insufficient diagnostic accuracy for estimating PMI, with a diagnostic accuracy of 0.6, which does not support our hypothesis. LE images (an equivalent of FFDM) provided a predicted suboptimal accuracy of 0.64. Adding DBT did not contribute to the diagnostic performance, lowering the accuracy to 0.55. RC demonstrated an accuracy of 0.67, which is slightly higher than LE, with the most accurate sign of PMI across all image types being muscle deformity and muscle enhancement on the MLO RC views. However, RC had the lowest reader confidence and poor inter-reader agreement- the worst among all image types. An example of an estimated true-positive CEM is presented in Fig. 5.

Fig. 5.

Pectoralis muscle invasion by MRI. 40-year-old woman with multicentric left breast cancer (invasive ductal, grade 3, ER+, PR+, HER2+, with DCIS). LE (5a), recombined (5b) and DBT slice (5c) images of left breast in craniocaudal projection demonstrated multiple masses (M in 5a-5c), with the extent of disease best visible on the RC image (5b). The contour of the pectoralis muscle is marked with arrowheads in 5a and 5c. There is enhancement outlining the contour of the pectoralis muscle (arrowheads in 5b). At least one of the masses overlaps with the contour of the pectoralis with no visible fat plane (arrows in 5a-5c). Four out of 6 radiologists (67 %) reported pectoralis muscle invasion on CEM with mean confidence of 2 out of 3. Post-contrast axial MRI image (5d) of the same patient demonstrated the known mass (M) extending beyond the contour of the pectoralis muscle (arrowheads) with focal muscle enhancement (arrow). 67 % of the radiologists reported muscle and fascia enhancement with a mean confidence of 2.7 out of 3. The patient underwent NAC with tumor response, and no muscle invasion was noted on final post-treatment surgical pathology.

Our preliminary study findings suggest that CEM may play a role in the evaluation of patients with subpectoral breast implants, in whom CEM incidentally performed better than MRI (Fig. 6). Only two cases with subpectoral implants were insufficient for a statistical analysis; however, in our study these cases were especially valuable due to the presence of SP-based ground truth. RC images of CEM correctly categorized the cases with and without PMI, while MRI had one true-positive and one false-positive result. Notably, the interpretative confidence of muscle enhancement on RC in the true-positive case was lower than in the true-negative case (1.5 vs 2.2). At the same time, MRI demonstrated muscle enhancement in both true-positive and false-positive cases, with reader confidence of 2.8 and 2.5, respectively.

Fig. 6.

Patient with subpectoral saline implants and CEM/MRI disagreement on PMI. 59-year-old woman with multifocal left breast cancer (invasive ductal, grade 1, ER, PR+, HER2+; DCIS, grade 2). LE (6a), recombined (6b) and DBT slice (6c) images of right breast in the mediolateral oblique projection demonstrated 2 masses (arrows). The contour of the pectoralis muscle is marked with arrowheads in 6a and 6c. Five of 6 radiologists (83 %) reported no muscle enhancement and no muscle invasion with confidence 2.3 out of 3. Post-contrast sagittal (6d) and axial (6e) MRI images of the same patient demonstrated the known malignant masses (arrows) directly abutting the pectoralis muscle with no intervening fat plane and with suggestion of muscle enhancement (arrowheads). Four of 6 radiologists (67 %) reported muscle or fascia enhancement with a confidence of 2.8 out of 3. The patient underwent total mastectomy with no NAC, with no PMI on pathology.

Our study is limited by the small number of cases, in particular, cases with SP-proven PMI. In addition, NAC administered to some patients necessitated using MRI as the proxy for the ground truth, which may diminish the validity of the conclusions. However, this difficulty reflects real-life practice due to the rare occurrence of “ideal” evaluable research cases with prepectoral masses visible on mammography and meeting all the imaging criteria, which would subsequently undergo surgical excision without NAC and demonstrate PMI. This scarcity of true positive cases is also likely to be prohibitive for acquiring sufficient reader experience for PMI evaluation on DBT/CEM in clinical practice.

Another limitation of our study is the variable experience of the readers with CEM. Yet this is of unclear consequence to the results, as there was no tendency to improve accuracy with increasing experience in breast imaging in general or specifically in CEM. Another observation of this study is that the reader confidence or inter-reader agreement did not directly correlate with accuracy. For example, DBT had the lowest accuracy but the highest confidence and inter-reader agreement.

Although breast MRI is the most accurate available imaging modality for the evaluation of prepectoral cancers, our study confirms its previously reported limitations. We demonstrated that breast MRI may not be reliable for PMI in the presence of perceived muscle enhancement (particularly in patients with subpectoral implants), despite high reader confidence. In contrast, some previous studies observed an opposite effect- the presence of PMI in the absence of muscle enhancement¹⁴.

In conclusion, combined DBT/CEM in our study did not provide reliable diagnostic information for predicting PMI, except in patients with subpectoral implants. The uncertainty of both MRI and DBT/CEM imaging signs limits the role of radiology in predicting PMI in equivocal cases beyond a statement that PMI cannot be excluded.

SEM Figure 1. Reader accuracy for PMI as a function of the number of CEM studies interpreted.

Supplementary Material

Supplementary material associated with this article can be found, in the online version, at doi:10.1067/j.cpradiol.2025.04.005.

Abbreviations:

BC

breast cancer

CEM

contrast-enhanced mammography

CC

craniocaudal projection

DBT

digital breast tomosynthesis

DBTCEM

combined DBT and CEM studies performed under the same compression

FFDM

full field digital mammography

LE

low energy image of CEM

MLO

mediolateral oblique projection

MRI

magnetic resonance imaging

NAC

neoadjuvant chemotherapy

PMI

pectoralis muscle invasion

RC

recombined image of CEM

US

ultrasound