Head-to-head Comparison of Perfluorobutane-Contrast Enhanced US and Multiparametric-MRI for Breast Cancer: A prospective, multicenter study

doi:10.21203/rs.3.rs-2191087/v1

Background: Multiparametric - Magnetic Resonance Imaging (MP-MRI) has high sensitivity for diagnosing breast cancers but cannot always be used as a routine diagnostic tool. To evaluate whether the diagnostic performance of Perfluorobutane (PFB)-contrast enhanced ultrasound (CEUS) is like MP-MRI in the breast cancer and if combining the two methods would enhance diagnostic efficiency.

Patients and Methods: This was a head-to-head, prospective, multicenter study. Patients with breast lesions diagnosed by US as Breast Imaging Reporting and Data System (BI-RADS) 3, 4, and 5 categories underwent both PFB-CEUS and MP-MRI scans. On-site operators and three reviewers categorized the BI-RADS of all lesions on two images, respectively. Logistic-bootstrap 1000-sample analysis and cross-validation were used to construct PFB-CEUS, MP-MRI, and hybrid (PFB-CEUS + MP-MRI) models to distinguish breast lesions.

Results: In total, 179 women with 186 breast lesions were evaluated from 17 centers in China. The area under the receiver operating characteristic curve (AUC) for the PFB-CEUS model (0.89; 95% confidence interval [CI]: 0.74, 0.97) was like that of the MP-MRI model (0.89; 95% CI: 0.73, 0.97), whereas both were inferior to the hybrid model (0.92, 95% CI: 0.77, 0.98). 90.3% false-positive and 66.7% false-negative results of PFB-CEUS radiologists, 90.5% false-positive and 42.8% false-negative results of MP-MRI radiologists could be corrected by Hybrid model, respectively. Dynamic nomograms of three models are accessible on websites.

Conclusions: PFB-CEUS can be used in differential diagnosis of breast cancer with the performance comparable to MP-MRI. Using PFB-CEUS and MP-MRI as joint diagnostics could further strengthen the diagnostic ability.

Trial registration: Clinicaltrials.gov; NCT04657328. Registered 26 September 2020.

PFB-CEUS

MP-MRI

Hybrid

diagnostic model

breast lesions

diagnostic performance

There exist more than twenty histological classifications for breast lesions. Diagnostic imaging for the identification of lesion features remains challenging due to the overlapping characteristics of some benign and malignant lesions(1,2). It is essential that imaging performance be optimized to meet the needs of patients with breast lesions.

Blood perfusion is one of the vital indicators in differentiating breast lesions (3). MP-MRI and CEUS are imaging methods to provide visualization of tumor blood perfusion (4). Initially, studies suggested that Multiparametric-MRI (MP-MRI) is more accurate than conventional ultrasound (US), or mammography(5,6), and could improve the DCE-MRI specificity especially(7). However, precautions must be taken in patients with renal dysfunction with respect to the use of MRI contrast agents(8). And cost, the timing of the MRI exam claustrophobia, patients with morbid obesity who cannot be accommodated within the MRI bore are the clinically relevant and commonly considered limitations of MRI(9).

Contrast enhanced ultrasound (CEUS) is a non-irradiating, accessible, and easy-to-implement imaging technique that is a powerful supplementary problem-solving tool in the context of MRI(10). However, the CEUS is not always recommended for routine clinical diagnostic use by several breast lesion management guidelines such as the American Cancer Society (ACS)(11), National Comprehensive Cancer Network (NCCN)(6), or the European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB)(12). The main reason cited is the lack of a specific pattern indicating malignancy due to the several retrospective studies showed the inconsistent results(Supplementary Table 1) (13–17).

Currently, sulfur hexafluoride (SHF) is the most widely available CEUS agent worldwide while with the inconsistent diagnostic results.

Image quality in the detection of breast cancer using SHF-CEUS is limited by insufficient signal generation from microbubbles released at high frequencies from linear transducers(18). Perfluorobutane (PFB) has been shown to have a high resonance frequency and stable outer shell and can resist acoustic pressure, leading to increased microbubble oscillations that minimize collapse and loss of signal(19–21). Therefore, PFB demonstrates a theoretical potential to improve the quality of the breast CEUS examination.

Therefore, we explored the application of feature analysis in a multicenter and larger patient cohort by comparing head-to-head PFB-CEUS and MP-MRI to investigate the difference in diagnostic ability between PFB-CEUS and MP-MRI for the identification of breast cancer and whether the combine PFB-CEUS can improve the MP-MRI diagnostic capacity of breast cancer.

Study Design

This study was a prospective, multicenter trial (ClinicalTrials.gov: NCT04657328) in which study participants were consecutively recruited from 17 centers in China from September 2020 to February 2021 (Supplementary Table 2). The inclusion criteria and exclusion criteria are shown in flowchart Figure 1 and Supplementary Materials and Methods S1. Twenty percent of the subjects were randomly selected to form a validation cohort and the remaining 80% was used as a development cohort. Institutional review board and regulatory approval were granted from each center, and all the patients provided written informed consent. All the investigators and authors had complete access to all the study results, and the authors had full control of the data and statistical results included in this report.

Data collection

Clinical characteristics and imaging features of PFB-CEUS and MP-MRI were prospectively collected in a Research Electronic Data Capture (REDCap) database.

Imaging Protocol

Radiologists at the study centers, each with more than ten years’ experience, performed the examination. PFB-CEUS and MP-MRI images from the 17 hospitals were stored in DICOM data format.

PFB-CEUS

The PFB-CEUS examinations were performed using 15 devices (Supplementary Table 3). With the linear probe, pulse inversion harmonic imaging and a mechanical index of 0.18 - 0.24 were used for PFB-CEUS. A bolus injection of 0.015ml/kg perfluorobutane-filled microbubble contrast agent (Sonazoid; GE Healthcare, Oslo, Norway) was administered via a ≧ 22-gauge catheter line placed in the antecubital vein. A 5-mL flush of 0.9% sodium chloride solution was administered after injection of the contrast agent. The dispersion was prepared just prior to use and administered within two hours of preparation. The imaging timer was started simultaneous to the completion of the contrast agent injection with continuous assessment of the lesions for one minute, followed by intermittent scanning for 10 seconds at the following time points: 1 min and 30 s, 2 min, 3 min, 4 min, and 5 min. Both the mass of interest and the breast involved were evaluated, and patients with multiple lesions were included based on the interval time between two injections being 20 minutes.

MP-MRI

The MP-MRI examinations were performed by using eight devices (Supplementary Table 4). All MP - MRI examinations were performed on 1.5 or 3.0 Tesla systems using a dedicated bilateral breast coil. Contrast material (gadolinium chelate) was administered as an intravenous bolus (maximum dose of 0.1 mmol per kilogram of body weight at 2 mL/sec), followed by a 20 mL saline flush. After contrast material administration, imaging was obtained at approximately 60 – 90 seconds. A small region of interest (ROI) (typically 3 x 3 x 1 voxels) in the lesion was used to measure the time-signal intensity curves.

Imaging Analysis

A panel of six radiologists (with each radiologist having ≧ 15 years’ experience for breast lesion diagnosis) reviewed 50 cases of PFB-CEUS images to identify and define the imaging characteristics. The MP-MRI imaging characteristics followed the ACR MRI BI-RADS guideline. Subsequently, two radiologists, each with a > 5 years’ experience, were trained with a ≧ 0.75 kappa value in all the images. The readers were blinded to current and previous breast imaging and histological findings. Discrepancies were resolved by consensus and consultation with one of the radiologists with 20 years’ experience.

Quantitative PFB-CEUS parameters of the lesions were acquired using quantitative analysis software, i.e., NovoUltrasound Kit (Precision Health Institute, GE Healthcare China) (Supplementary Figure1). ROI 1 included the entire tumor boundary on PFB-CEUS imaging, while avoiding the surrounding parenchyma. ROI 2 encircled the normal-appearing parenchyma, including the same image acquisition plane as far as possible from the breast tumor(22).

Reference Standard

Histopathology was used as the reference standard. Tissue samples of the lesion were obtained by US-guided biopsy or surgical resection. Biopsy was performed using a 14/16-gauge core needle with real-time US/PFB-CEUS guidance. Specimens were reviewed by two senior pathologists from each academic practice.

Statistical Analysis

To detect a difference of 0.1 between a diagnostic test using PFB-CEUS, with an area under the ROC curve (AUC) of 0.85 and using MRI with an AUC of 0.95, a sample size of 81 malignant and 55 benign lesions were needed to achieve 90% power at a significance level of 0.05. Sample size calculations were performed using PASS version 11 (NCSS, LLC, Kaysville, Utah, USA).

The univariate and multivariable logistic regression analysis was performed to select the risk factors for breast cancer diagnosis and construct three different models: PFB-CEUS model, MP-MRI model and Hybrid model. Each model contained clinical features (Age, Diameter, Breast density, Menopausal status, Childless) and radiological features.

Radiological features of PFB-CEUS contained (fall time (FT), rise time (RT), time-to-peak (TTP), mean transit time (mTT), arrival time (AT), derivation of washing time (earlier, later, or synchronous), degree of enhancement (hyperenhancement, iso-enhancement, or hypo-enhancement), uptake pattern (centripetal, centrifugal, diffuse, or no enhancement), presence or absence of entire washout time > 5 minutes, heterogeneous pattern, rim-like enhancement, claw-shaped pattern, perfusion defects, size enlargement, poorly-defined margin, irregular size, and nourishing vessels)).

Radiological features of MP-MRI contained (background parenchymal enhancement (minimal, mild, moderate, or marked), enhancement type, type of lesion (focus/foci only, mass, or non-mass enhancement), uptake pattern (centripetal, centrifugal, diffuse, or other), kinetics—initial phase (slow, medium, or rapid), kinetics—delayed phase (persistent, plateau, or washout), signal intensity of T1 and T2, (high, equal, or low), DWI (high or low), and ADC value according to the Breast Imaging Reporting and Data System MRI lexicon)).

Radiological features of Hybrid model contained all the radiological features of PFB-CEUS and MP-MRI.

Model selection was performed under stepwise criteria, when necessary. Bootstraps of 1000 resamples and five-fold and ten-fold cross validation were performed to evaluate the performance of the models (the bootstrap process is shown in Supplementary Materials and Methods S2(23). Nomograms are shown free . Discrimination was quantified by using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. Calibration was assessed with the Hosmer-Lemeshow test and calibration plot. Descriptive analysis summarized the patient characteristics. The operating point selection method is shown in Supplementary Materials and Methods S3. All analyses were performed using R and Stata (version 15). P values less than 0.05 were considered statistically significant.

Participant and Imaging Characteristics

In total, 186 lesions from 179 female participants (mean age, 48 ± 11 years) with 115 malignant lesions and 71 benign lesions were enrolled after excluding 177 lesions (Figure 1) from 17 tertiary centers across China, allowing for the inclusion of a diverse patient population. The mean age (SD) of the overall cohort was 49 ± 11 years, and all 179 of the patients were women (Table 1). The details of overall pathologic distribution and of each cohort are presented in Supplementary Table 5. The frequencies of the PFB-CEUS and MP-MRI characteristics in the two data cohorts are described in Supplementary Table 6.

Model Development and Performance

The univariate and multivariable logistic regression analysis of PFB-CEUS, MP-MRI, and the Hybrid model is shown in Supplementary Table 7, 8, and 9 respectively.

In the external validation model, The hybrid model showed an AUC greater than that of both MP-MRI and PFB-CEUS and with AUC 0.92 ([95% CI: 0.77, 0.98]).the PFB-CEUS model was shown to have an AUC 0.89 ([95% CI: 0.74,0.97]), which was better than that with the PFB-CEUS BI-RADS (AUC 0.80 [95% CI: 0.63,0.92), P =.43; Luo models (AUC 0.74 [95% CI: 0.57,0.87], P =.01)(24); Chen model: (AUC 0.69 [95% CI: 0.51, 0.83], P =.02)(25); and Yukio model (AUC 0.70 [95% CI:0.53, 0.85], P =.01)(13). In addition, The MP-MRI model was shown to have an AUC 0.89 ([95% CI: 0.73, 0.97]), which was better than that with the ACR BI-RADS models (AUC 0.73 [95% CI: 0.55, 0.87], P =.15). (Supplementary Table 10, Figure2A).

The PFB-CEUS, MP-MRI and hybrid models were well-calibrated and all showed statistical significance (P >.05) in the Hosmer-Lemeshow test. The calibration plot are showed in Figure 2B. The decision curve are showed in Supplement Figure 2. In addition, the respective dynamic nomograms of the MP-MRI, PFB-CEUS, and hybrid models are shown in the following links: https://ceus.shinyapps.io/PFB-CEUS/ , https://ceus.shinyapps.io/MP-MRI/, https://ceus.shinyapps.io/Hybrid/. Figure 3A shows a matched example in which the PFB-CEUS and MP-MRI nomograms showed a high malignant probability of breast cancer. Figure 3B shows a matched example in which the PFB-CEUS showed a high malignant probability of cancer but MP-MRI nomograms showed a low malignant probability of cancer which the lesion was diagnosed as invasive carcinoma by pathology. Figure 3C showed a matched example in which the PFB-CEUS and MP-MRI nomograms showed a high malignant probability of cancer but Hybrid nomograms showed a low malignant probability of cancer which the lesion was diagnosed as hyperplasia by pathology.

Comparison among the three models

PFB-CEUS vs. MP-MRI model

The PFB-CEUS model showed a comparable discrimination ability for diagnosing breast cancer, with the MP-MRI model showing the same not only in the development cohort ((AUC 0.90 [95% CI: 0.84, 0.94]) vs. (AUC 0.90 [95% CI: 0.85, 0.95], P =.80) but also in the validation cohort (AUC 0.89 [95% CI:0.74, 0.97]) vs. (AUC 0.89 [95% CI: 0.73, 0.97], P =.85) (Supplementary Table 10).

PFB-CEUS vs. Hybrid model

The hybrid model had a higher capacity to diagnose breast cancer compared with the PFB-CEUS model ((AUC 0.95 [95% CI: 0.90, 0.98]) vs. ((AUC 0.90, [95% CI: 0.84,0.94]), not only in the development cohort; P =.01) but also in the validation cohort ((AUC 0.92 [95% CI: 0.77, 0.98]) vs. (AUC 0.89 [95% CI: 0.74, 0.97]), P =.29) (Supplementary Table 10).

MP-MRI vs. Hybrid model

The hybrid model demonstrated a higher capacity for diagnosing breast cancer compared with the MP-MRI model ((AUC 0.95 [95% CI: 0.90, 0.98]) vs. (AUC 0.90 [95% CI: 0.85, 0.95]), respectively; P =.078), not only in the development cohort but also in the validation cohort ((AUC, 0.92 [95% CI: 0.77, 0.98]) vs. (AUC 0.89 [95% CI: 0.73,0.97]), respectively; P =.401) (Supplementary Table 10).

Model performance in sub-populations

Of all the subgroups dichotomized by age and menstrual status, the hybrid model demonstrated a higher AUC for diagnosis than the PFB-CEUS model for patients younger than 49 years of age ((AUC, 0.96 [95% CI: 0.89, 0.99]) vs. (AUC 0.86 [95% CI: 0.76,0.93]), respectively) and premenopausal ((AUC, 0.94 [95% CI: 0.87,0.98]) vs. (AUC 0.86 [95% CI: 0.77,0.92]). No significant difference in AUC was noted between other comparisons of modality (Supplementary Table 11). The diagnostic results for lesions with high-risk malignancy (a lesion that would be appropriate for surgical consultation) including intraductal papilloma, hyperplasia, phyllodes tumor are presented in Supplementary Table 12.

Model performance for diagnosing breast cancer compared with radiologists

BI-RADS 4A, 4B, and 4C were used as the cut point for comparison.

PFB-CEUS

When compared with the on-site radiologists, the PFB-CEUS model achieved higher sensitivity at the BI-RADS 4B+ and 4C+ modes with lower specificity at all modes, as well as the hybrid model. When compared with the three senior reviewers, the PFB-CEUS model achieved higher sensitivity at the BI-RADS 4B+ and 4C+ modes, with lower specificity at all three modes. However, the hybrid model achieved higher sensitivity at all three BI-RADS modes, and higher specificity at the BI-RADS 4C+ mode (Table 2).

MP-MRI

When compared with the on-site radiologists, the MP-MRI model achieved higher sensitivity at all three modes with lower specificity at all three modes, as well as the hybrid model. When compared with the three senior reviewers, the MP-MRI model achieved higher sensitivity at all three modes, higher specificity at the BI-RADS 4C+ mode. The hybrid model achieved higher sensitivity at all three BI-RADS modes, and higher specificity at the BI-RADS 4B+ and 4C+ mode (Table 2).

Model Performance with False-positive and False-negative Correction Rate

PFB-CEUS

The false-positive correction rate of the PFB-CEUS and hybrid models was 80.6% and 90.3% for on-site results, and 82.2% and 88.9% for the reviewers’ results based at the BI-RADS 4A+ modes, respectively. The false-negative correction rate of the PFB-CEUS and Hybrid model was 66.7% and 66.7% for on-site results, and 83.3% and 83.3% for reviewers’ results at the BI-RADS 4A+ modes, respectively (Supplementary Table 13) (Figure 4).

MP-MRI

The false-positive correction rate of the MP-MRI and Hybrid model was 77.7% and 86.1% for on-site results, and 83.0% and 90.5% for reviewers’ results at the BI-RADS 4A+ modes, respectively. The false-negative correction rate of the MP-MRI and Hybrid models was 50.0% and 0.0% for on-site results, and 57.1% and 42.8% for reviewers’ results at the BI-RADS 4A+ modes, respectively (Supplementary Table 13) (Figure 4).

MP-MRI has better sensitivity and specificity for the detection of breast cancer than dynamic contrast-enhanced MRI (DCE-MRI)(26). However, this technology was not always available and was associated with prohibitively high costs for use as a routine diagnostic tool(27). There is an increasing demand to develop a surrogate, easy-to-implement method to diagnose breast lesions. Our study is the first to compare the diagnostic performance between PFB-CEUS and MP-MRI, both advanced breast cancer diagnosis modalities, for the diagnosis of breast cancer.

In this study, the AUC for the PFB-CEUS model was like that of the MP-MRI model, but both were inferior to the hybrid model. For the subgroup of patients, the hybrid model demonstrated a higher AUC than the PFB-CEUS models for patients older than 50 years of age and premenopausal for diagnosis. The hybrid model could improve the sensitivity and specificity of the PFB-CEUS on-site radiologists in all three modes, as well as the MP-MRI. The FPCR and FNCR was excellent for the PFB-CEUS, MP-MRI, and the hybrid model. The results indicate PFB-CEUS could potentially improve the diagnostic ability of MP-MRI for breast cancer, and when patients were restricted by the contraindications of MP-MRI, PFB-CEUS could perform as an alternative examination method for patients with breast lesions.

Over the past decade, several comparative studies to investigate the ability of SHF-CEUS and DCE-MRI for breast cancer diagnosis have been published, and these have demonstrated inconsistent results. (Supplementary Table 1). Half indicated that SHF-CEUS showed higher sensitivity and specificity than DCE-MRI and the remainder presented contrasting results. In all these studies, there was no rigorous imaging feature review or selection, and a lack of evaluation of interobserver variability between radiologists and inclusion analysis of quantitative SHF-CEUS indicators. Compared with the previous retrospective cohort study between SHF-CEUS and DCE-MRI for breast cancer, it is worth mentioning that PFB-CEUS and MP-MRI images were prospectively collected according to the standard protocol in our study.

Notably, our study has confirmed the theoretical advantages of PFB which had a high resonance frequency and stable outer shell, leading to the depiction of clear enhancement signals such as tumor size and duration and improving the CEUS image quality in linear transducers. Enhancement size enlargement and FT became stable indicators through a bootstrap analysis that excluded the zero value, insuring a more accurate estimation of PFB-CEUS. Compared with the previously constructed CEUS model, our study showed a higher AUC than all the others in both the development cohort and validation cohorts (14,24,25). In addition, MP-MRI was set as the control with PFB-CEUS in the present study, which can reflect the diffusion of water molecules in the intracellular and extracellular spaces and can lead to a higher specificity than DCE-MRI in the diagnosis of malignant lesions due to the high cell density and the small extracellular space in malignant lesions(26). Encouragingly, PFB-CEUS didn’t sacrifice the diagnostic performance for breast cancer compared to MP-MRI with its superior contrast resolution.

When PFB was used as an agent to compare with MRI for breast cancer, only one result was reported, namely, that PFB-CEUS reached a similar AUC compared with DCE-MRI in 127 patients(14). Our study involved patients who were prospectively enrolled from 17 tertiary centers across China, providing for a head-to-head comparative study with a relatively large sample size that guaranteed the generalization results. Rigorous validation including the 1000-sample bootstrap, five- and ten-fold internal validation and independent external validation, ensured robust and reproducible results, even when using different devices settings, and the comprehensive features were analyzed for the ability to distinguish breast cancer by using clinical parameters, and qualitative and quantitative analysis of perinodular and intranodular features selected from PFB-CEUS and MP-MRI.

Compared with the radiologists, the implementation of all three models were feasible in daily practice, requiring less time and effort to collect and analyze clinical and imaging characteristics (only 3 - 5 variables). Free- nomograms with relatively reliable results are structured for simple daily practice, accessible to three different websites.

Furthermore, a false-positive imaging diagnosis may lead to unnecessary biopsy invasion, patient anxiety, and health care costs. BI-RADS 4A was usually recommended for use in the clinic as the cut-off grade for biopsy. In our study, the FPCR of PFB-CEUS, MP-MRI, and the hybrid model was excellent, particularly when BI-RADS 4A was used as the cut-off point to diagnose positive lesions, indicating that the models’ FPCR could supply a reference to appropriately avoid overdiagnosis for probable benign lesions.

Our study had several limitations. First, the absolute accuracy and stability of the models may have been influenced by the relatively small sample size, and the lesion size and breast density failed to perform statistical analysis , although this was calculated in the design of this prospective study. Second, because we enrolled only lesions with BI-RADS grades 3, 4, and 5 in the grey-scale US, the prevalence of breast cancer in these participants maybe be higher than that found during the usual diagnostic examination for all surveillance-positive lesions. Therefore, the diagnostic performance of PFB-CEUS and MP-MRI is potentially biased.

In conclusion, PFB-CEUS model was at least as efficient and accurate as that of MP-MRI model. Using PFB-CEUS and MP-MRI as joint diagnostics could could further strengthen the ability to better characterizing breast cancer. This suggests that not only can PFB-CEUS be used as a surrogate method to diagnose breast lesions, but also that when these models are used as an aid to radiologists in clinical practice, they can save time and effort and avoid overdiagnosis.

Funding: This work was supported by Grants 82030047, 92159305 from the National Scientific Foundation Committee of China.

Correspondence to:

Jie Yu, mailing address: Department of Interventional Ultrasound, Fifth Medical Center of Chinese PLA General Hospital,China, 28 Fuxing Road, Beijing 100853, China; phone numbers:010-66939530; email address: [email protected].

Conflicts of interest: There are no conflicts of interest.

Ethical Approval and consent to participate

The study was approved in the center of the principal investigator at Chinese PLA General Hospital in Beijing, China(IRB number: 2020-300). The Ethical Committee was “Chinese PLA General Hospital”. The approval was obtained on July 23th, 2020. Every patient signed a written informed consent previously approved by ethical committee. The research was carried out in accordance with the Declaration of Helsinki.

Authors' contributions

Manlin Lang and Jie Yu wrote the main manuscript text. Manlin Lang, Ping Liang, Huiming Shen, Hang Li, Ning Yang, Bo Chen, Yixu Chen, Hong Ding, Weiping Yang, Xiaohui Ji, Ping Zhou, ligang Cui, Jiandong Wang, Wentong Xu, Xiuqin Ye, Zhixing Liu, Yu Yang, Tianci Wei, Hui Wang, Yuanyuan Yan, Changjun Wu, Yiyun Wu, Jingwen Shi, Yaxi Wang, Xiuxia Fang, Ran li collected the data. All authors reviewed the manuscript.

Availability of data and materials

Data generated or analyzed during the study are available from the corresponding author by request.

Tosteson ANA, Fryback DG, Hammond CS, Hanna LG, Grove MR, Brown M, et al. Consequences of false-positive screening mammograms. JAMA Intern Med. 2014;174:954–61.
den Dekker BM, Bakker MF, de Lange SV, Veldhuis WB, van Diest PJ, Duvivier KM, et al. Reducing False-Positive Screening MRI Rate in Women with Extremely Dense Breasts Using Prediction Models Based on Data from the DENSE Trial. Radiology. 2021;301:283–92.
Folkman J. What is the evidence that tumors are angiogenesis dependent? J Natl Cancer Inst. 1990;82:4–6.
Sridharan A, Eisenbrey JR, Dave JK, Forsberg F. Quantitative Nonlinear Contrast-Enhanced Ultrasound of the Breast. AJR Am J Roentgenol. 2016;207:274–81.
Riedl CC, Luft N, Bernhart C, Weber M, Bernathova M, Tea M-KM, et al. Triple-modality screening trial for familial breast cancer underlines the importance of magnetic resonance imaging and questions the role of mammography and ultrasound regardless of patient mutation status, age, and breast density. J Clin Oncol. 2015;33:1128–35.
Bevers TB, Helvie M, Bonaccio E, Calhoun KE, Daly MB, Farrar WB, et al. Breast Cancer Screening and Diagnosis, Version 3.2018, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 2018;16:1362–89.
Clauser P, Krug B, Bickel H, Dietzel M, Pinker K, Neuhaus V-F, et al. Diffusion-weighted Imaging Allows for Downgrading MR BI-RADS 4 Lesions in Contrast-enhanced MRI of the Breast to Avoid Unnecessary Biopsy. Clin Cancer Res. 2021;27:1941–8.
Contrast Manual | American College of Radiology [Internet]. [cited 2022 Sep 8]. Available from: https://www.acr.org/Clinical-Resources/Contrast-Manual
Mann RM, Cho N, Moy L. Breast MRI: State of the Art. Radiology. 2019;292:520–36.
M A, Hj J, Tk K, K K. Contrast-enhanced US of the Liver and Kidney: A Problem-solving Modality. Radiology [Internet]. Radiology; 2022 [cited 2022 Sep 9];303. Available from: https://pubmed.ncbi.nlm.nih.gov/35191740/
Saslow D, Boetes C, Burke W, Harms S, Leach MO, Lehman CD, et al. American Cancer Society guidelines for breast screening with MRI as an adjunct to mammography. CA Cancer J Clin. 2007;57:75–89.
Sidhu PS, Cantisani V, Dietrich CF, Gilja OH, Saftoiu A, Bartels E, et al. The EFSUMB Guidelines and Recommendations for the Clinical Practice of Contrast-Enhanced Ultrasound (CEUS) in Non-Hepatic Applications: Update 2017 (Long Version). Ultraschall Med. 2018;39:e2–44.
Du J, Wang L, Wan C-F, Hua J, Fang H, Chen J, et al. Differentiating benign from malignant solid breast lesions: combined utility of conventional ultrasound and contrast-enhanced ultrasound in comparison with magnetic resonance imaging. Eur J Radiol. 2012;81:3890–9.
Miyamoto Y, Ito T, Takada E, Omoto K, Hirai T, Moriyasu F. Efficacy of sonazoid (perflubutane) for contrast-enhanced ultrasound in the differentiation of focal breast lesions: phase 3 multicenter clinical trial. AJR Am J Roentgenol. 2014;202:W400-407.
Does contrast-enhanced ultrasound (CEUS) play a better role in diagnosis of breast lesions with calcification? A comparison with MRI - PubMed [Internet]. [cited 2022 Sep 8]. Available from: https://pubmed.ncbi.nlm.nih.gov/32463295/
Li C, Yao M, Shao S, Li X, Li G, Wu R. Diagnostic efficacy of contrast-enhanced ultrasound for breast lesions of different sizes: a comparative study with magnetic resonance imaging. Br J Radiol. 2020;93:20190932.
Liu W, Zong M, Gong H-Y, Ling L-J, Ye X-H, Wang S, et al. Comparison of Diagnostic Efficacy Between Contrast-Enhanced Ultrasound and DCE-MRI for Mass- and Non-Mass-Like Enhancement Types in Breast Lesions. Cancer Manag Res. 2020;12:13567–78.
Sun C, Sboros V, Butler MB, Moran CM. In vitro acoustic characterization of three phospholipid ultrasound contrast agents from 12 to 43 MHz. Ultrasound Med Biol. 2014;40:541–50.
Boca Bene I, Dudea SM, Ciurea AI. Contrast-Enhanced Ultrasonography in the Diagnosis and Treatment Modulation of Breast Cancer. J Pers Med. 2021;11:81.
Sontum PC. Physicochemical characteristics of Sonazoid, a new contrast agent for ultrasound imaging. Ultrasound Med Biol. 2008;34:824–33.
Piskunowicz M, Sridharan A, Poznick L, Silvestro E, Hwang M. Optimization of Mechanical Indices for Clinical Contrast-Enhanced Ultrasound. J Ultrasound Med. 2021;40:1963–70.
Lee YJ, Kim SH, Kang BJ, Kim YJ. Contrast-Enhanced Ultrasound for Early Prediction of Response of Breast Cancer to Neoadjuvant Chemotherapy. Ultraschall Med. 2019;40:194–204.
Clinical Prediction Models | SpringerLink [Internet]. [cited 2022 Sep 8]. Available from: https://link.springer.com/book/10.1007/978-3-030-16399-0
Breast contrast-enhanced ultrasound: is a scoring system feasible? A preliminary study in China - PubMed [Internet]. [cited 2022 Sep 8]. Available from: https://pubmed.ncbi.nlm.nih.gov/25133534/
Luo J, Chen J-D, Chen Q, Yue L-X, Zhou G, Lan C, et al. Predictive model for contrast-enhanced ultrasound of the breast: Is it feasible in malignant risk assessment of breast imaging reporting and data system 4 lesions? World J Radiol. 2016;8:600–9.
Zhu C-R, Chen K-Y, Li P, Xia Z-Y, Wang B. Accuracy of multiparametric MRI in distinguishing the breast malignant lesions from benign lesions: a meta-analysis. Acta Radiol. 2021;62:1290–7.
Pan I-W, Oeffinger KC, Shih Y-CT. Cost-Sharing and Out-of-Pocket Cost for Women Who Received MRI for Breast Cancer Screening. J Natl Cancer Inst. 2022;114:254–62.

TABLE 1. Baseline Characteristics of the Study Patients.

Characteristic	Overall (n=186)	Development cohort (n=151)	Validation cohort (n=35)
Age (y)^†	49 ± 11 (19, 76)	49 ± 11 (19, 76)	49 ± 10 (27, 68)
Age (y)^†
< 49	90 (48.4)	76 (50.3)	14 (40.0)
≧49	96 (51.6)	75 (49.7)	21 (60.0)
Body mass index (kg/m²)^†	23.9 ± 3.5 (13.7, 37.0)	23.9 ± 3.3 (16.4, 37.0)	23.5 ± 4.2 (13.7, 34.1)
Body mass index (kg/m²)
< 18.5	6 (3.4)	3 (2.1)	3 (9.1)
18.5 - < 25	110 (61.5)	91 (62.3)	19 (57.6)
≥ 25.0	63 (35.2)	52 (35.6)	11 (33.3)
Dense breast
YES	137 (76.5)	107 (73.3)	30 (90.9)
NO	42 (23.5)	39 (26.7)	3 (9.1)
Parity status
Nulliparous	16 (8.9)	14 (9.6)	2 (6.1)
Parous	163 (91.1)	132 (90.4)	31 (93.9)
Menopausal status^‡
Premenopausal	107 (59.8)	89 (61.0)	18 (54.5)
Postmenopausal	72 (40.2)	57 (39.0)	15 (45.5)
First-degree relatives with breast cancer§
Presence	169 (94.4)	138 (94.5)	31 (93.9)
Absence	10 (5.6)	8 (5.5)	2 (6.1)
Diagnosis method
Surgery	73 (39.2)	62 (41.1)	11 (31.4)
Core biopsy	113 (60.8)	89 (58.9)	24 (68.6)
Lesion size (cm)^*	1.8 (1.2, 2.4)	1.8 (1.2, 2.5)	1.8 (1.3, 2.4)
Lesion size (cm)^*
< 1.5	65 (34.9)	52 (34.4)	13 (37.1)
≧ 1.5	121 (65.1)	99 (65.6)	22 (62.9)
Histologic type
Benign	71 (38.1)	58 (38.4)	13 (37.1)
Malignant	115 (61.8)	93 (61.6)	22 (62.9)

Note.—Unless otherwise indicated, data are numbers of patients, and data in parentheses are percentages.

^* Data are medians, and data in parentheses are the interquartile range.

^†Data are means ± standard deviation, and data in parentheses are the range.

^‡ Menopausal status: Women aged 60 years or older, reporting a history of hysterectomy, or reporting no periods within the past 12 months

without use of hormonal contraceptives were categorized as postmenopausal. Women reporting regular periods (12–18 times in the last 12

months) without use of hormonal contraceptives were categorized as premenopausal.

^§ First-degree relatives (mother, sister, and daughter) with breast cancer.

TABLE 2. Comparison of breast cancer diagnostic performance of each model with radiologists in different BI-RADS modes.

	BI-RADS (3 vs. 4a+)		BI-RADS (3, 4a vs. 4b+)		BI-RADS (3, 4a, 4b vs. 4c+)
	Se%	Sp%	Se%	Sp%	Se%	Sp%
PFB-CEUS
On-site^*	97.4 (112/115) (92.6, 99.5)	56.3 (40/71) (44.0, 68.1)	93.9 (108/115) (87.9, 97.5)	78.9 (56/71) (67.6, 87.7)	78.3 (90/115) (69.6, 85.4)	91.5 (65/71) (82.5, 96.8)
Reviewers^†	94.8 (109/115) (89.0, 98.1)	36.6 (26/71) (25.5, 48.9)	86.1 (99/115) (78.4, 91.8)	70.4 (50/71) (58.4, 80.7)	78.3 (90/115) (69.6, 85.4)	78.9 (56/71) (67.6, 87.7)
CEUS model	NA	NA	97.4 (112/115) (92.6, 99.5)	36.6 (26/71) (25.5, 48.9)	88.7 (102/115) (81.4, 93.8)	80.3 (57/71) (69.1, 88.8)
Hybrid model	100.0 (115/115) (96.8, 100.0)	25.4 (18/71) (15.8, 37.1)	100.0 (115/115) (96.8,100.0)	50.7 (36/71) (38.6, 62.8)	90.4 (104/115) (83.5, 95.1)	84.5 (60/71) (74.0, 92.0)
MP-MRI
On-site^*	98.3 (113/115) (93.9, 99.8)	49.3 (35/71) (37.2, 61.4)	92.2 (106/115) (85.7, 96.4)	73.2 (52/71) (61.4, 83.1)	80.0 (92/115) (71.5, 86.9)	87.3 (62/71) (77.3, 94.0)
Reviewers^†	93.9 (108/115) (87.9, 97.5)	25.4 (18/71) (15.8, 37.1)	89.6 (103/115) (82.5, 94.5)	46.5 (33/71) (34.5, 58.7)	81.7 (94/115) (73.5, 88.3)	67.6 (48/71) (55.5, 78.2)
MRI model	100.0 (115/115) (96.8, 100.0)	11.3 (8/71) (5.0, 21.0)	98.3 (113/115) (93.9, 99.8)	40.8 (29/71) (29.3,53.2)	88.7 (102/115) (81.4, 93.8)	80.3 (57/71) (69.1, 88.8)
Hybrid model	100.0 (115/115) (96.8, 100.0)	25.4 (18/71) (15.8, 37.1)	100.0 (115/115) (96.8, 100.0)	50.7 (36/71) (38.6, 62.8)	90.4 (104/115) (83.5, 95.1)	84.5 (60/71) (74.0, 92.0)

Note.—The model’s sensitivity was determined by setting the model’s specificity to match the on-site/reviewers’ specificity; model’s specificity was determined by setting the model’s sensitivity to match the on-site/reviewers’ sensitivity. Se = Sensitivity, Sp = Specificity. NA means the PFB-CEUS model’s breast lesion diagnosis as BI-RADS 3 grade is absent. Data in parentheses are the numerator and denominator and the 95% confidence intervals.

^* The radiologists who performed PFB-CEUS and MP-MRI.

^† The radiologists who reviewed the stored PFB-CEUS and MP-MRI images.

No competing interests reported.

BCRSupplementarymaterials.docx

Head-to-head Comparison of Perfluorobutane-Contrast Enhanced US and Multiparametric-MRI for Breast Cancer: A prospective, multicenter study

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Materials and Methods

Results

Discussion

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1