Contrast-enhanced Mammographic Features of In- situ and Invasive Ductal Carcinoma Manifesting Microcalci cations Only: Feasibility for Predicting Underestimation

Yun-chung Cheung (  alex2143@cgmh.org.tw ) Linkou Chang Gung Memorial Hospital: Chang Gung Medical Foundation https://orcid.org/0000-00026870-3429 Kueian Chen Chang Gung Memorial Hospital Linkou Main Branch: Chang Gung Memorial Hospital Chi-Chang Yu Chang Gung Memorial Hospital Linkou Main Branch: Chang Gung Memorial Hospital Shir-Hwa Ueng Chang Gung Memorial Hospital Linkou Main Branch: Chang Gung Memorial Hospital Chia-Wei Lee General Electric Company Shin-Cheh Chen Chang Gung Memorial Hospital Linkou Main Branch: Chang Gung Memorial Hospital


Background
The manifestation of suspicious microcalci cations only on mammography is indicative of early cancers. Approximately 20%-25% of these patients are diagnosed with malignancy by mammographyguided needle biopsy [1][2][3]. Most of the cases are noninvasive ductal carcinoma in situ (DCIS); however, unfortunately, invasive components occasionally embed within the noninvasive cancer bed, which will alter preoperative planning. Regardless of DCIS or invasive ductal carcinoma (IDC), it is universally agreed that management with subsequent surgical treatment either with conservative or total mastectomy should be recommended [4][5][6]. However, being able to predict the underestimation of biopsyproven DCIS will facilitate preoperative planning, in which it is not essentially advised to perform sentinel lymph node sampling for cases of pure DCIS [7]. Obviation of the supplementary performance of sentinel lymph node biopsy will provide bene ts, including shortening the operative time, avoiding unnecessary exposure to radiation doses or minimizing the potential complications of lymph node resection.
Mammography-guided vacuum-assisted core needle biopsy is commonly used to histologically diagnose the aetiology of suspicious microcalci cations. The performance has improved from spring-loaded biopsy to vacuum-assisted large needle biopsy, which approximates surgical biopsy [8]. However, DCIS underestimation by needle biopsy occasionally occurs in clinical practice. A meta-analysis of 7350 cases of DCIS including masses or microcalci cations from 52 studies reported a 30.3% underestimation rate for 14-gauge core needles and an 18.9% underestimation rate for 11-gauge vacuum-assisted needle biopsy [9]. Even when using a large-bore 7-gauge biopsy needle, the upgrade rate of DCIS to IDC is still 15.38% [10].
Multiple imaging modalities, including conventional mammography and sonography, have been used in an attempt to predict DCIS underestimation [11][12][13][14]. In this study, we investigate a new imaging technique, contrast-enhanced spectral mammography (CESM), which can provide conventional low-energy mammograms (LMs) and recombined enhanced images (REIs) from the same session of a single compressed position. With this exclusive bene t, concerning enhancements on REIs and suspicions on conventional mammograms can be easily correlated and assessed. In the past, research mainly focused on the diagnostic performance of CESM. In this retrospective study, we compared the enhanced features of biopsy-diagnosed breast cancers, including DCIS and IDC, manifesting mammographic microcalci cations only on conventional mammograms to analyse the feasibility of predicting DCIS underestimation preoperatively.

Patient selection
With approval from the IRB of our hospital, we reviewed patients who underwent mammography-guided biopsy, had suspicious breast microcalci cations and underwent CESM within 2 weeks before the biopsy.
The patients were selected according to the following criteria: (1) mammographic microcalci cations only; (2) mammography-guided biopsy proven cancer (including DCIS and IDC); and (3) subsequent surgical treatment in our hospital.
Patients with (1) abnormal renal function (abnormal serum creatinine >1.0 mg/dL and glomerular ltration rate 60 mL/min/1.73 m 2 ), (2) pregnancy, (3) lactation, (4) allergy history to iodized contrast medium, (5) past history of breast cancer or surgery before biopsy or (6) systemic disease such as hyperthyroidism were clinically excluded from receiving CESM examination. The CESM examination method and potential side effects of contrast medium were fully explained to those who underwent CESM. Finally, all the patients signed agreements to participate in this study according to our hospital regulations.

Image analysis
The CESM (Senographe Essential or Pristina CESM; GE Healthcare, Buc, France) examination was standardized and performed with intermittent exposure (approximately 2-second intervals) to low and high energy (below and above the k-edge of iodine: 33.2 keV) during a single breast-compressed position. The image acquisitions were routinely obtained in the sequence of craniocaudal (CC) and mediolateral oblique (MLO) views of the bilateral breasts within 2-6 min after the start of a single-bolus injection of nonionic contrast medium (Omnipaque 350 mg I/mL; GE Healthcare, Dublin, Ireland) at a rate of 3 mL/s for a total dose of 1.5 mL/kg body weight via an intravenous catheter that was inserted into the forearm prior to the examination.
All the CESM data, including the LMs and REIs, were reviewed. The locations of the biopsied microcalci cations were rst identi ed on the LM and then evaluated for the enhancement features on the REI. The presence or absence of enhancement was rst recorded, and enhancement features, including size, morphology and texture, were evaluated. The size of cancer enhancement was measured in the greatest diameter in either the CC or MLO view. The enhancement morphology was classi ed as nonmass (clump appearance without a bulging outline) or mass (shaped appearance with a bulging outline). The textures of enhancement consisted of pure ground glass (transparent to underlying) and unpuri ed ground glass (with a nontransparent solid part). The appearances were demonstrated ( Fig. 1,

Statistical Analysis
We used the Mann-Whitney U test for statistical analysis of the signi cance of enhanced features between DCIS and IDC. The signi cance was then evaluated by univariable logistic regression for diagnostic performance. All statistical analyses were performed with SPSS software version 20.0 (SPSS, Chicago, IL, USA). We also used the Wilcoxon rank sum test to analyse the signi cant differences among the three groups of DCIS, upgraded IDC and IDC. A p-value of <0.05 was set to indicate statistical signi cance.

Patient characteristics
From 2015 to 2020, 56 patients with breast cancer (44 DCIS and 12 IDC) diagnosed by stereotactic-or tomographic-guided vacuum-assisted core needle biopsy were identi ed from our records. However, only 49 breast cancer patients (39 DCIS and 10 IDC) were enrolled for analysis after excluding 7 patients without subsequent operation in our hospital. Of the 39 biopsy-diagnosed DCIS cases, 12 (30.77%) were surgically/histologically upgraded to IDC. Finally, 27 patients with DCIS and 22 patients with IDC were compared in this study.
The average ages of patients with DCIS and IDC were approximately the same (53.9 years vs 51.4 years). The morphologies of biopsied microcalci cations on mammography were recorded as amorphous in 15 cases, pleomorphic in 24, linear in 7 and casting in 3; and the distributions were 27 in group, 7 in region, 13 in segment and 2 in linear ( Table 1). All the microcalci cations were nally classi ed into category 4 of the American College of Radiology Breast Imaging and Reporting Data System (ACR BI-RADS) after assessment and recommended for mammography-guided vacuum-assisted core needle biopsy.  The presence or absence of enhancement and the enhanced texture were signi cantly different between the two groups. Univariate logistic regression further showed moderate diagnostic performance in the former (p value < 0.01, AUC = 0.66, sensitivity = 93%, speci city = 39%) and in the latter (p value < 0.01, AUC = 0.74, sensitivity = 0.65, speci city = 83%).

Statistical analysis of the DCIS, upgraded IDC and IDC groups
The Wilcoxon rank sum test was used to analyse the signi cant differences among three groups of DCIS (27 cases), upgraded IDC (12 cases) and IDC (10 cases). The DCIS group showed a signi cant difference in morphology compared with the IDC group (p = 0.0134) and in enhanced texture compared with the upgraded IDC and IDC groups (p = 0.0226 and 0.0018, respectively). These results are shown in Fig. 5.

Discussion
In this study, the new modality we investigated, CESM, is a novel mammography-based imaging examination that was approved for clinical use in 2011 [15]. Utilizing the different attenuation coe cients of iodine and glandular tissues under low-and high-energy exposures, computer software can recombine the low-and high-energy images to highlight the presence of iodine accumulation after eliminating the breast tissue background. The enhanced lesions indicate possible pathogenic lesions. The sensitivity of CESM ranges from 93-100%, and the speci city ranges from 63-88%, showing signi cant improvement compared to full-eld digital mammography [16-18]. Particularly for dense breasts, CESM bene cially increased the cancer sensitivity and speci city by approximately 22% and 16%, respectively [19].
Although mammography is sensitive for microcalci cation detection, the cancer probability ranges widely among the various morphologies or distributions of microcalci cations on mammograms, from 2-95% in accordance with ACR BI-RADS 4 [3]. By using an enhancing technique, advanced contrast-enhanced magnetic resonance imaging (CE-MRI) can detect the associated enhancement of glandular tissues adjacent to suspicious microcalci cations. A meta-analysis analysed 1843 lesions from 20 studies and reported a pooled sensitivity and speci city of 92% and 82%, respectively, for BI-RADS 4 microcalci cations [20]. CESM has an approximate performance to CE-MRI, with 88.89% sensitivity, 86.56% speci city, 72.72% positive predictive value and 95.08% negative predictive value from a screening population [21]. Compared to CE-MRI, CESM more easily correlates the microcalci cations on the LM to surrounding glandular enhancement on the REI in the same session of positioning. Another test was designed to have radiologists read the LM rst and then read the LM with the REI a day later. The sensitivity, speci city, positive predictive value and negative predictive value of CESM were mildly improved from LM, from 93.8-96.8%, from 36.6-34.1%, from 54-54% and from 88.2-92.2%, respectively [22]. Conversely, the enhancement information from REI did not seem to affect the decision in terms of the bias of personal mammographic knowledge on microcalci cations. Another investigation recently reported a signi cantly higher predictive positive value and lower misdiagnosis rate when using REIs than when using LMs alone, and a machine learning model could signi cantly improve the diagnosis of both low-risk (AUC 0.77 using LMs alone to 0.9 with REIs) and high-risk (AUC 0.71 using LMs alone to 0.86 with REIs) groups of microcalci cations [23]. The additional REI seemed helpful in assessing suspicious microcalci cations.
For the management of suspicious microcalci cations, mammography-guided vacuum-assisted needle biopsy, either with stereotactic or tomosynthesis techniques, is clinically used as a standard procedure to obtain calci ed specimens for pathologic diagnosis. Once microcalci cations are diagnosed as cancer, subsequent surgery should be performed to remove the residual cancer cells. However, whether sentinel lymph node sampling is performed depends on the actual nature of pure DCIS or IDC.
Due to the limited amount of specimens obtained by needle biopsy, there is a risk of underestimation. In this series, the DCIS underestimation rate was 30.77% (12 of 39 biopsy-diagnosed DCIS cases). To gain more knowledge for preoperative prediction, we analysed information on cancer enhancement from CESM, including the presence or absence of enhancement and the enhancement extent, morphology and texture of cancers. From our results, all the unenhanced microcalci cations were DCIS, and all the IDC cases showed enhancement. The incidence of DCIS upgrades seemed rare in cases of unenhanced needle-diagnosed DCIS manifesting only microcalci cations. This result was speculated to be due to the silent or less aggressive behaviour of DCIS.
The preoperative measurement of cancer extent to predict whether a coexisting invasive component exists may serve as a guideline for surgeons to consider sentinel lymph node sampling [7]. Larger cancers tend to have a greater chance of having invasive components. The cut-off sizes of DCIS underestimation varied, ranging from 1.1 cm to 4 cm [24,25.26], which was unfortunately variable in different patient collections. In our CESM study, the average size of IDC (2.93 cm) was larger than that of DCIS (1.46 cm), but the difference was statistically nonsigni cant.
In terms of morphology, using mass or nonmass descriptions to characterize a cancer is theoretically related to the growth pattern of cancer. The mass morphology is de ned as a localized lesion with a bulging outline con guration, and the nonmass morphology is an in ltrating pattern with a clump appearance. In our study, the incidence of nonmass enhancement was approximately 59.1% for DCIS and 40.9% for IDC. More IDC cases presented mass enhancement than DCIS cases (72.2% vs 27.8%). In fact, similar results from previous CE-MRI studies have shown that the presence of a mass lesion was a preoperative predictor of DCIS with an invasive component [27].
Stronger enhancement secondary to richer neovascularity indicates the rapid or aggressive behaviour of a cancer. Such enhancement that could obscure the underlying tissue was herein described as solid enhancement. In this study, we used the term 'ground glass' to describe the texture of mild enhancement with transparent visualization to the background; otherwise, in the cases of unpuri ed texture, we indicated the presence of nontransparent solid enhancement. Unpuri ed enhancement was predominantly observed in IDC compared to DCIS, accounting for 83.3% vs 16.7% of cases.
There are several limitations in this study.
(1) The case number was small. CESM was not a compulsory examination prior to biopsy in clinical practice, and certain patients hesitated to undergo contrast medium injection. (2) The biopsy performance could technically in uence the underestimation. Specimen radiography could lead to the retrieval of microcalci cations for diagnosis but was unable to avoid underestimation. (3) The underestimation absolutely depends on microscopic ndings. The biopsy target and whether invasive parts are obtained for microscopic evaluation are key to avoiding underestimation. The obtained microcalci cations in the specimens were not equivalent in terms of the presence of invasive elements. Other than for the purpose of diagnosis, enhancement features such as mass and solid enhancements on CESM could provide suggestive sites for biopsy that may improve the diagnosis of invasion. The usefulness of CESM-guided biopsy should be proven in the future.