Conspicuity and imaging features of breast cancers on digital breast tomosynthesis according to molecular prole

Objective To describe BI-RADS features and evaluate conspicuity of breast cancer on digital breast tomosynthesis (DBT) according to their molecular prole. method Institutional review board was obtained for this retrospective study. Consecutive patients with histologically proven breast cancers who underwent digital breast tomosynthesis (DBT) with 2D synthetic mammography (SM) and digital mammography (DM) at the time of diagnosis were included. Morphological features and conspicuity of cancers on DM, SM and DBT were evaluated in consensus by two breast radiologists. Imaging features were compared across molecular subtypes (luminal, triple negative (TN) and HER2+) using Fisher’s exact test and between DBT and SM and DM using McNemar’s test. Conspicuity was compared between DBT and SM and DM using Wilcoxon matched pairs test and between each molecular subtype using the Wilcoxon test.


Introduction
According to its genetic pro le, breast cancer is a heterogeneous disease with variable histological and clinical features [1]. Genetic characteristics of breast cancers are re ected by histological features such as tumor subtypes, grade, proliferation index (Ki67) and hormonal receptor expression [2]. The World Health Organization (WHO) has de ned different types of cancers associated with variable prognosis according to hormonal receptor expression: Luminal, with expression of Estrogen and/or Progesterone receptor expression; Non-luminal HER-2+ (Human Epidermal Growth Factor Receptor-2) with no hormonal receptor expression and overexpression of HER-2; and Triple negative (TN) cancers with no hormonal receptor nor HER-2 overexpression [3]. It has been widely reported that the different subtypes of breast cancers may demonstrate speci c 2D mammographic features. In particular, luminal cancers typically present as a mass with irregular shape and spiculated margins on 2D mammography, while TN cancers are more likely to demonstrate oval shape and microlobulated margins or present as asymmetries on mammography and they can mimic benign lesions [4]. In order to avoid misinterpreting such types of subtle malignant lesions in the clinical setting, it is crucial that radiologists are aware of the speci c features of the different breast cancer subtypes on imaging. Although correlation between histology and imaging has been widely described previously on digital mammography (DM), speci c features of breast cancers on digital breast tomosynthesis (DBT) have rarely been reported. Only one recent study suggested that morphological features on DBT could be associated with molecular characteristics [5], but very little data are available to support this assertion. Breast cancer features on 2D synthetic mammography (SM) have not yet been reported, although it has been reported that the additional cancers detected with the adjunct of DBT in the screening setting would preferentially be less aggressive tumors with luminal molecular pro les [6,7]. However, other recent studies found no such a difference [8,9]. It is therefore of interest to evaluate whether molecular features have an impact on conspicuity of breast cancers on DBT, which has, to the best of our knowledge, not been evaluated previously.
The objective of this study is to evaluate whether the different molecular subtypes of invasive breast cancers are associated with speci c morphological features on DBT and SM. The second objective is to evaluate the association between molecular pro les and conspicuity of breast cancers on DBT and SM in comparison with DM.

Material And Methods
Study population IRB approval was obtained for this retrospective study and informed consent was waived.
Consecutive patients with invasive breast cancer diagnosed between March 2018 and November 2019 who underwent DBT with SM at our institution of at least the affected breast (at least one view, medio lateral oblique or cranio caudal) and a recent (< 3 months) DM of at least the affected breast at the time of diagnosis were eligible for inclusion. Patients were ineligible if: presence of a post-biopsy marker clip on DM or DBT/SM images, personal history of prior surgery of the ipsilateral breast and absence of invasive carcinoma in the surgical specimen.

Imaging acquisition
In our institution, DBT (Selenia dimension, Hologic, Bedford, Mass.) is routinely performed as a complementary imaging examination to DM in the diagnostic setting for characterization purposes and as part of the pre-treatment workup of histologically proven breast cancers. In most cases it is bilaterally in two views, in some cases unilaterally [10].
DM were performed at least for the affected breast, in at least one view [Cranio-Caudal (CC) or Medio lateral Oblique (MLO)], either at our institution or outside our institution. DM performed outside our institution were either downloaded into our local Picture Archiving and Communication System (PACS) (Entreprise Imagine®, Agfa Healthcare), or reviewed on lms.

Image analysis
All DBT and SM images were reviewed on dedicated workstations (Secure 3D, Hologic). DM images were either reviewed on dedicated workstations (55/111) or on lms (55/111). Two breast radiologists (*BLINDED* and *BLINDED*) with fellowship level training and respectively 20 and 14 years of experience in breast imaging and 11 and 6 years of experience in DBT, read all examinations in consensus. Readers were blinded to the molecular pro le of the lesions and tumor locations. DM, SM and DBT were reviewed in three distinct reading sessions each separated by 4 weeks, to reduce potential recall bias. In situations where multiple lesions were present, instructions were given to the readers to assess the most prominent one (index tumor).

Location and conspicuity
Readers rst assessed the location and conspicuity of the tumor. Conspicuity was de ned as the combination of the con dence in the presence of a lesion with the con dence in decision-making based on lesion detectability. Conspicuity was assessed on the following ve-point scale: 0: no visible ndings, 1: low conspicuity, 2: medium conspicuity, 3: high conspicuity and 4: very high conspicuity [11,12]. In cases where readers determined a wrong location of the tumor (side and quadrant), conspicuity was reported as "non-visible".

Characterization
For DM and SM sessions, lesion features were assessed according to the 2013 edition of the BI-RADS lexicon [13]. Description of abnormalities on DBT is not de ned in the BI-RADS lexicon. However, as routinely performed in our institution, readers were asked to describe lesions on DBT according to terms used in the fth edition of the BI-RADS lexicon [14]. In particular, the term "mass" was assigned to lesions clearly identi ed and occupying a volume whereas "asymmetry" was assigned to lesions that had concave borders, contained fat or demonstrated low conspicuity. When only one view was available, no term was used for characterization of asymmetries (i.e. focal and global); and since no previous examination was available for the readers the term "developing" was not used. Finally, with each technique, readers were instructed to measure the maximum diameter of the index tumor.
Pathology and assessment of tumor size When available, tumor sub-type, grade, HER-2 status, Estrogen and Progesterone receptors, KI67 and presence of in situ ductal carcinoma and tumor size were determined from pathology reports of the nal surgical specimen. In patients who underwent neo-adjuvant chemotherapy (NAC) or had no surgery, tumor sub-type was determined from ultrasound-guided 14-Gauge core needle or 8-Gauge stereotacticguided vacuum-assisted biopsies. For assessment of tumor size, tumor maximum diameter on pathology of the surgical specimen was considered the gold standard. In cases of multifocal breast cancer, the maximum diameter of the largest invasive tumor (index tumor) was assessed.

Statistical analysis
Proportion of morphological features of breast cancers was compared between the different molecular pro les: Luminal, TN and HER-2 + cancers using Fisher's exact test.
Qualitiative features on each technique were compared using McNemar's test.
Conspicuity of breast cancers was compared between the different tumor subtypes, molecular characteristics, low (< 20%) and high (> 20%) Ki67, and grade using the Wilcoxon test. Conspicuity between DM, SM and DBT was compared using Wilcoxon matched pairs tests. P values less than 0.05 were considered statistically signi cant. Statistical analysis was performed using SAS V9.4 software.

Study population and patient characteristics
A total of 136 patients were eligible for inclusion in our study. Thirteen patients were excluded because of presence of a post biopsy-marker clip and 6 because of history of prior surgery of the ipsilateral breast. Six patients had only in situ carcinoma on the surgical specimen and were also excluded. A total of 111 patients comprised our study population.
Mean patient age was 64 ± 13 years (Table 1). Seventy-eight (70%) cancers were palpable at the time of diagnosis and 33 were not palpable (30%). On pathology, our population consisted of 87 invasive carcinoma of non-speci c type (IC-NST), 13 invasive lobular carcinomas (ILC), four mixed invasive lobular and non-speci c type carcinomas, four papillary, two apocrine and one mucinous carcinomas.    SM than on DM (P < .0001 and P = 0.0004 respectively) and on DBT than on SM (P = 0.003).

Conspicuity
Breast cancers were signi cantly more conspicuous on DBT than on DM (P = 0.03) but differences between DM and SM and SM and DBT did not reach signi cance (P = 0.2 and P = 0.25 respectively). Eleven cancers were not visible on DM, 13 on SM and 6 on DBT. Among the eleven cancers not visible on DM, two cancers were also not visible on SM and DBT (2 ICNST grade 3, one luminal and one HER-2+).
Non-palpable breast cancers were signi cantly more conspicuous on DBT than on DM (P = .02) whereas no differences were found for palpable cancers between the two techniques (P = 0.47). Cancers with low proliferation indices were less conspicuous on DM than tumors with KI67 ≥ 20% (P = .008), whereas this difference was not signi cant on DBT (P = .11).

Discussion
Our results show that breast cancers may demonstrate different features on DBT than on DM. In particular, cancer more often demonstrates architectural distortion on DBT. Moreover, we showed that according to their molecular pro le, breast cancers could demonstrate speci c features on DBT and on SM. In our study, TN breast cancers more frequently presented as masses with microlobulated margins on DBT and, like on DM, HER-2 positive tumors more frequently demonstrated microcalci cations on DBT and on SM. We also observed that conspicuity of luminal breast cancers was better on DBT than on DM while conspicuity of HER-2 + or TN tumors was not signi cantly improved.
Breast cancer is a clinically, histologically and molecularly heterogeneous disease with variable prognosis. It is now well established that according to their histological features, breast cancers may demonstrate different morphological features on digital mammography. Hence, it has been reported that luminal subtype breast tumors more likely present as spiculated masses than triple negative cancers [15].
It has also been observed that HER-2 + tumors were more frequently associated with microcalci cations on DM than other subtypes [16]. Being aware of the speci c features of the different histological types of breast cancers on imaging is of great importance in clinical practice. In particular, knowing that some cancers can more often demonstrate less suspicious features on DBT, such as microlobulated margins for TN tumors, may incite breast radiologists to pay particular attention when diagnosing a breast lesion with such characteristics. Moreover, at the time when more and more publications suggest that mathematical parameters derived from radiomic analysis of breast cancers might be interesting biomarkers of cancer aggressiveness and prognosis, it is of interest to show that tumor morphological features on a recent technique such as DBT may also re ect the underlying pathology [17][18][19].
Since it is a relatively recent technique, speci c features of breast cancers on DBT have indeed been scarcely described. To the best of our knowledge, only one study evaluated the appearance of breast cancers on DBT according to their histology and molecular pro le [5]. In line with our results, they showed that presence of calci cations within the tumors on DBT was more frequent in HER-2 + than in other types of breast cancers. However, our study is the rst to show that triple-negative tumors may also demonstrate speci c feature on DBT i.e. presenting more frequently as masses with microlobulated margins than other cancers. Moreover, we also showed that DBT could predominantly improve conspicuity of less aggressive breast cancers. Improvement of conspicuity of breast cancers on DBT has also been shown to vary according to pathology and was more pronounced for lobular than for ductal invasive carcinomas [12]. Although we could not demonstrate such a difference in our study, probably because of the relatively small number of lobular carcinomas in our population, our nding are in line with previous publications, suggesting that additional breast cancers detected on DBT would be more predominantly luminal subtypes and low-grade tumors [7,20]. This increased detection of luminal subtypes cancers on DBT compared to DM might be due to their speci c features, i.e. more frequent spiculated margins, that are known to be particularly well visualized on tomosynthesis [21]. In this context, our study provides interesting results with regards to the histological features of breast cancers underlying their conspicuity on DBT in comparison with DM.
Our study has some limits. First it is a retrospective study with a relatively small sample size, in particular for the number of TN and HER-2 + tumors. Secondly, in our study TN and HER-2 + were more often palpable that luminal cancers and we did not evaluate whether better conspicuity on DM was due to bigger tumor size or to speci c morphological features on imaging. Further studies including more cases of TN and HER-2 + breast cancers would be needed to determine whether speci c features on imaging are independent factors of better conspicuity of breast tumors on DBT. Finally, 50% of DM cases were reviewed on lms and not on the dedicated workstation which could have altered conspicuity and characterization assessments with DM, although some publications reported that soft-copy reading provides no advantage for interpretation [22,23].

Conclusion
We showed that breast cancers may demonstrate speci c features on DBT according to their molecular features. Our result suggest that DBT may predominantly improves the conspicuity of less aggressive tumors, which would bene t the most from an increased detection rate with the use of DBT.
Declarations the right breast, presenting as a mass with microlobulated margins; (C) 69-year-old woman with invasive carcinoma of non-speci c type of the right breast luminal B, presenting as a mass with spiculated margins.