Comparative study of multiple breast cancer screening methods in the evaluation of breast non-mass-like lesions

Objective: To compare multiple breast cancer screening methods for evaluating breast non-mass-like lesions (NMLs) and to investigate the best screening method for breast non-mass-like lesions (NMLs). Methods: This retrospective study examined 253 patients aged 24 to 68 years who were diagnosed with breast NMLs from April 2017 to December 2019. All lesions were evaluated by MG, HHUS, and ABUS to determine BI-RADS classication, underwent pathological examination within six months or at least 2 years of follow-up. The sensitivity, specicity, accuracy, positive predictive values (PPV), and negative predictive values (NPV) of MG, HHUS, and ABUS features in the prediction of malignancy were compared. Independent risk factors for malignancy were assessed using non-conditional logistic regression. Results: MG, HHUS, and ABUS ndings signicantly differed between benign and malignant breast NML, including internal echo, hyperechoic spot, peripheral blood ow, internal blood ow, catheter change, peripheral change, coronal features of ABUS, and structural distortion, asymmetry, and calcication in MG. ABUS was superior to MG and HHUS in sensitivity, specicity, PPV, NPV, as well as in evaluating the necessity of biopsy and accuracy in identifying malignancy. MG was superior to HHUS in specicity, PPV, and accuracy in evaluating the need for biopsy. HHUS was distinctly superior to MG in sensitivity and NPV in determining malignancy; however, specicity, PPV, and accuracy were similar. Moreover, internal blood ow, calcication, and coronal plane feature were independent risk factors in distinguishing benign and malignant lesions. Conclusions: ABUS was superior to HHUS and MG in evaluating the need for biopsy and differentiating benign and malignant in breast NMLs. Compared to each other, HHUS and MG had their own relative advantages. Internal blood ow, calcication, and coronal plane feature were independent risk factors in identifying benign and malignant lesions. 4B. Pathology: Intraductal carcinoma.


Introduction
Mammographic screening, as an effective method for early detection of the breast cancer, can reduce mortality rate of breast cancer [1]. Ultrasound and automated breast ultrasonography (ABUS) are considered to be important complementary means for breast cancer screening by mammography, especially dense breast [2][3][4][5][6]. However, at least 25% of detectable breast cancer would be missed by Mammographic screening [7], including some cases of non-mass-like lesions (NMLs). Although mammographic screening cannot diagnosis breast non-mass-like lesions (NMLs), it can detect the features of NMLs which include structural distortion, asymmetry and calci cation. On ultrasonography (US) , NMLs refer to lesions that have no clear boundaries ,no spatial mass effect on two or more different scanning directions [8].NMLs account for 9.21% of all breast abnormalities [9], and lesions that show ambiguous hypoechoic characteristics and architectural distortion without duct-like structure are usually de ned as NML [10].
Ultrasonography and mammography can detect mass-like abnormalities as standardized by the American College of Radiology Breast Imaging Reporting and Data System atlas (BI-RADS) [11]. Currently, breast BI-RADS classi cation using ultrasound is not considered for detection of NML, and there are no standard NML guidelines. Recently, studies based on B-mode, color Doppler ow imaging (CDFI), and strain or shear-wave elastography to distinguish benign from malignant NMLs have demonstrated their diagnostic performance. Similar to non-mass-like enhancement on breast MRI, with the enhanced image quality of high-resolution ultrasonography and widespread use of bilateral whole-breast examinations, it is possible to detect the hypoechoic regions (NMLs) that do not meet criteria for space-occupying lesions de ned by BI-RADS [12,13]. NMLs re ect a wide range of breast lesions, including benign, high-risk, and malignant lesions. High-risk lesions presenting as an NML include atypical ductal hyperplasia, lobular carcinoma in situ, papillary tumor, and squamous epithelial atypia. Squamous epithelial atypia is observed in NMLs, although absent squamous epithelial atypia has been reported [13].
In breast ultrasound, our understanding of the speci c pathological features that predict malignancy is incomplete. It is true that the methods used to evaluate breast US ndings vary across practices and have been somewhat intuitive. In addition, there have been few studies on the relationship between US, NML lesion and their pathological importance [10,14]. Ko KH et al. [10] classi ed breast NMLs into four categories based on calci cation and architectural distortions, which are correlated with different BI-RADS categories. However, overlapping features make it di cult to accurately categorize breast NMLs, so tissue biopsy is required for classi cation. Thus, ultrasound features are helpful in reducing biopsies of benign lesions [13].
Automated breast ultrasonography, also known as automated breast volume scanning, is a new imaging technology that can provide standardized image acquisition and coronal images of the entire breast. This system received FDA clearance as an auxiliary means to screening mammography in 2008 [14]. By examining the breast continuously in transverse sections, ABUS can automatically perform threedimensional breast reconstruction and simultaneously obtain morphologic and coronal images [6]. ABUS enhances the sensitivity, speci city and accuracy of breast lesion discrimination [15]. However, few studies have examined the value of hand-held ultrasonography (HHUS) or ABUS in diagnosing breast NMLs. As we all know, different screening methods have their own advantages. However, there are few studies on the relationship among different screening methods, NML lesion ndings and their pathological signi cances. This study compared and analyzed the evaluation of breast NMLs by three different breast cancer screening methods: mammography, HHUS, and ABUS. Our ndings will make contributions to selecting effective screening methods for the detection of NMLs, and assist in reducing the incidence of misdiagnosis and NML biopsy rate.

Participants
This retrospective study examined 253 patients aged 24 to 68 years who were diagnosed with breast NMLs from April 2017 to December 2019. All patients underwent MG, HHUS, and ABUS before breast surgery or biopsy. We excluded those who did not undergo all 3 imaging studies or pathological examination or did not receive at least 2 years follow-up. Patients who had previously undergone surgery and pathological examination were also excluded.
Only patients diagnosed with breast NML were enrolled. NMLs included ill-de ned geographic hypoechoic or clustered hyperechoic spots without mass and tubular hypoechoic duct-like structural or architectural distortions on HHUS. These lesions exhibited suspicious calci cation, distorted structure, asymmetry, and dilated ducts on mammography. The study was carried out in accordance with the Declaration of Helsinki and approved by the Ethical committee of the rst a liated hospital, Jinan University (ZE2020-232).
NML was rst detected by ultrasonography in 17 patients younger than 35 years of age who had a family history of cancer or a personal high risk of cancer. Two hundred and forty patients also underwent surgery and pathology within six months. Thirteen patients had benign NMLs, of which 5 were diagnosed according to breast MRI criteria (BI-RADS category 1, 3 cases; BI-RADS category 3, 2 cases) and 8 were The NML was located in the central region of the ultrasonography image, and two-dimensional longitudinal and transverse sections and CDFI were stored. All NML features were evaluated and recorded, including location, maximum diameter, echo pattern, structural distortion, ductal changes, microcalci cation (hyperechoic, <2 mm in diameter [16]), and posterior echo. To describe the distribution of microcalci cation more accurately, scattered and aggregated point hyperechoic were used.
All NMLs were classi ed according to BI-RADS categories [8]. CDFI was evaluated according to Adler's grade. The category in two-dimensional sonography was used as the reference for ABUS classi cation. However, if the coronal appearance of ABUS was consistent with mass, the lesions were classi ed according to the lexicon of ACR BI-RADS. All mammography and MRI feature were evaluated using the lexicon of ACR BI-RADS.

Statistical analysis
Quantitative data are expressed as means with standard deviation and categorical data as composition or rate ratios. Quantitative data were compared using the t-test or Kruskal-Wallis test as appropriate.
Categorical data were compared using the chi-square test or Fisher's exact test. In uencing factors were analyzed by unconditional stepwise logistic regression. Sensitivity, speci city, and accuracy of mammography, HHUS, and ABUS for differentiating breast NMs were calculated using nal pathologic ndings as the reference. Statistical analyses were performed using SPSS software version 22.0 (Chicago, USA). P <0.05 was considered signi cant.

Results
Pathologic diagnosis of breast NML Among the 253 study patients, 73 (28.9%) had lesions classi ed as malignant or precursor and 180 (71.1%) as benign. The pathologic diagnoses are presented in Table 1 years of follow-up (Fig. 2). Clustered punctate calci cations (hyperechoic punctate) were observed in 16 patients. The maximum lesion diameter was <10 mm in 68 patients and >20 mm in 64. Association between imaging features and pathological diagnosis As shown in Table 2, there were signi cant differences between benign and malignant breast NMLs in the following characteristics: hyperechoicity, peripheral change, ductal changes, microcalci cation, posterior echo, peripheral CDFI, internal CDFI, coronal plane feature (ABUS), calci cation (MG), and structural (MG).
These ndings indicate that these imaging features acquired by different techniques can effectively predict pathological diagnosis. Indicative capability of different imaging techniques to evaluate the necessity of biopsy.
As shown in Table 3 ABUS was superior in sensitivity, speci city, PPV, NPV, and accuracy in evaluating the need for biopsy compared to MG and HHUS. MG was superior to HHUS in speci city, PPV, and accuracy; HHUS was superior to MG in sensitivity and NPV. ABUS was superior in sensitivity, speci city, PPV, NPV, and accuracy in determining malignant breast NML compared to MG and HHUS. HHUS was distinctly superior to MG in sensitivity and NPV but similar in speci city, PPV, and accuracy.   [18]. Although MG has been shown to reduce the incidence of breast cancer, its effect on biopsy rate is an ongoing debate [19]. In this study, we found that ABUS had the highest diagnostic sensitivity, speci city, PPV, NPV and accuracy. MG was superior to HHUS in evaluating the need for biopsy in speci city, PPV, and accuracy. However, in terms of determining malignancy, HHUS was superior to MG in sensitivity and NPV, but similar in speci city, PPV, and accuracy.
Conventional ultrasonography characteristics such as internal echo, hyperechoic spot, peripheral blood ow, internal blood ow, catheter changes, peripheral changes, as well as the coronal plane of ABUS are signi cant factors in the evaluation of breast NMLs and provide a theoretical basis for accurate ultrasonographic differentiation of benign and malignant lesions. MG is important in screening and diagnosing breast cancer. Its main aim is to identify densities, microcalci cations, and asymmetry [20].
Ultrasonography can assist in further characterization. Breast densities may be solid or cystic with smooth or irregular margins [21]. Microcalci cations may be focal or diffuse, and calci cations may be coarse or ne. Multifocal, ne calci cations are more likely to be malignant, whereas uniform, large, coarse calci cations are usually benign. Furthermore, they may be stable or increase over time [22], which is characteristic of NMLs. Proliferative NMLs with calci cation are often classi ed as more dangerous and require biopsy. In this study, NMLs in 64 patients exhibited hyperplasia, including 43 cases of hyperplasia with calci cation and 3 of hyperplasia with apocrine metaplasia and calci cation. Among these, 23 were classi ed as more dangerous and biopsy was recommended.
Breast symmetry is often the most di cult to characterize, as it shows great variation between individuals as well as between the left and right breasts of the same individual and even between different breast quadrants. As the primary screening tool for breast cancer, the sensitivity of conventional MG is approximately 70% [23], and the sensitivity decreases with the increase of the quality of breast tissue assessed. Moreover, 76% of cancers are missed in women with dense breast tissue, while the overlap of normal breast tissue can lead to false-positives [24,25].
ABUS allows non-invasive imaging of tissue using real-time sonography with high sensitivity, speci city, and accuracy [26]. At different strain levels, ABUS acquires different sonographic features of fat, normal glandular tissue, brous tissue, DCIS, and in ltrating ductal carcinoma [27]. Moreover, vascular distribution by CDFI provides data regarding blood supply [28]. CDFI can depict microvascularity and allows continuous dynamic observation of microcirculation. In our study, abundant blood ow was more frequently observed in malignant NMLs, similar to existing studies [10]. These features allow ABUS to differentiate various mass lesions. In comparisons of HHUS and ABUS evaluation of breast NML malignancy, most studies have reported higher sensitivity for ABUS but higher speci city and accuracy for HHUS [29], which is consistent with our ndings. ABUS can focus on a lesion from three orthogonal sections at the same time and display the accurate spatial position of the lesion in real-time, which allows a thorough and detailed evaluation [30].
In previous studies, lower diagnostic sensitivity and speci city restricted the use of ABUS [31]. Therefore, it is critical to determine independent breast NML risk factors to enhance diagnostic capability. The logistic regression model that combined NML and internal ow and coronal surfaces detected by ABUS provided a more objective and accurate method for characterizing NML. However, age is also important (breast lumps), as shown by epidemiology [32].
There are several limitations to our study. First, the differences between observers and the standardization of image storage are well-known imaging limitations. To reduce these, all imaging was retrospectively analyzed by two experienced experts while standardizing the scanning. The concise de nition of descriptors was discussed, and an evidence-based consensus was reached. However, it remains necessary to standardize an accurate and speci c de nition of NML. Second, the study was conducted in a single institution and involved a small number of subjects, so the disease distribution was not necessarily generally representative. Further large-scale multicenter studies are warranted.
In conclusion, conventional ultrasonographic (HHUS) characteristics, such as internal echo, hyperechoic spot, peripheral blood ow, internal blood ow, catheter change, and peripheral change, as well as coronal features (ABUS), structural distortion (MG), asymmetry (MG) and calci cation characteristics (MG) are signi cant in evaluating the risk of malignancy and need for biopsy in breast NMLs. Internal blood ow (HHUS), calci cation (MG), and coronal features (ABUS) may be risk factors for malignant or precursor lesions. ABUS evaluation signi cantly increased the diagnostic sensitivity, speci city, PPV, NPV, and accuracy. HHUS was distinctly superior to MG in sensitivity and NPV in determining malignancy; however, MG was superior in speci city, PPV and accuracy in evaluating the need for biopsy.

Declarations
Ethic approval and consent to participate Nightingale rose graph of division of cases