A convenient model based on mammography and magnetic resonance imaging for preoperative differentiation of breast phyllodes tumors and fibroadenomas

Introduction

Fibroepithelial tumors of the breast are biphasic tumors composed of epithelial and stromal proliferation, mainly including fibroadenomas (FAs) and phyllodes tumors (PTs) (1). Clinically, FAs and PTs often present as painless masses, and it is difficult to distinguish them. However, there are significant differences in the treatment plans. FAs are usually managed with clinical follow-up and require only enucleation (2), while PTs are typically treated with surgical excision with wide margins and supplemented with postoperative radiation or chemotherapy due to their invasiveness and tendency to recur (3-6). Preoperative differentiation of FAs and PTs can not only avoid unnecessary surgeries but also help to develop personalized treatment plans to improve patient prognosis. Therefore, the differentiation of FAs and PTs has been frequently evaluated in clinical research.

Core needle biopsy is an important method for the preoperative differentiation of FAs and PTs (7). Although PTs exhibit distinct intracellular growth patterns, characteristic phyllodes stroma, high mitotic activity, and high stromal cell atypia under the microscope (1,8), the limited tissue samples obtained from biopsies still show overlapping histological features of epithelial and stromal proliferation in both FAs and PTs (9,10). This makes the differentiation between the two more challenging, especially between FAs and benign PTs (2,5,11). Additionally, in both FAs and PTs, there are mutations in the MED12 gene (8,12-14), and other molecular features, such as RARA mutation (15), SETD2, MAP3K1, etc., can contribute to differentiation but lack specificity (16). Moreover, the immunophenotypic characteristics often overlap (3), particularly in the presence of CD34 staining in stromal cells of both FAs and PTs (17).

Currently, imaging examinations play an increasingly important role in the diagnosis of breast diseases. Ultrasonography, mammography (MG), and magnetic resonance imaging (MRI) are the three commonly used imaging modalities for breast examinations. Previous studies have confirmed that these imaging methods and their features are helpful in differentiating FAs and PTs (18,19). However, due to the low incidence of PTs, only a small number of PTs were included in previous studies, and there were few studies in which two imaging modalities were combined, resulting in insufficient analysis of the imaging features. Therefore, the aim of this study is to compare the morphological characteristics of PTs and FAs based on MG and magnetic resonance (MR) images so as to facilitate the development of a convenient tool for the preoperative differentiation of the two lesions. We present this article in accordance with the TRIPOD reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-145/rc).

Methods

Patients

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study is a retrospective analysis that solely utilized fully anonymized data retrieved from the hospital information system. The Institutional Review Board of Fudan University Shanghai Cancer Center confirmed exemption from ethical approval and waived the requirement for informed consent. FA group: the clinical and imaging data of patients who underwent breast imaging examinations from October 2019 to December 2020 at Fudan University Shanghai Cancer Center were retrospectively collected. PT group: the clinical and imaging data of patients who underwent breast imaging examinations from January 2011 to December 2020 were retrospectively collected. Inclusion criteria were following: (I) female patients; (II) preoperative MG and breast MR images; (III) masses visible on both MG and MR images; (IV) pathological diagnosis of FA/PT based on postoperative biopsy, with clear pathological classification; (V) available complete clinical and imaging data. Exclusion criteria were following: (I) history of treatment or biopsy before imaging examination; (II) poor image quality, poor visualization of masses, or inability to obtain time-intensity curve (TIC); and (III) multiple masses with different pathological results. Finally, a total of 147 FA patients and 138 PT patients were included in the study, including 50 benign PTs, 58 borderline PTs, and 30 malignant PTs (Figure 1).

Figure 1 Flowchart of the inclusion of fibroadenoma and phyllodes tumor patients. FA, fibroadenoma; MG, mammography; MR, magnetic resonance; PT, phyllodes tumor; TIC, time-intensity curve.

Breast MG and MRI examination

All patients underwent routine craniocaudal (CC) and mediolateral oblique (MLO) MG. The examination equipment included digital MG machines from GE (USA), Siemens (Germany), and Hologic (USA). After the images were taken, they were transmitted to Fudan University Shanghai Cancer Center PACS workstation.

The patient’s MRI examination was conducted with two 1.5T and two 3T scanners. The imaging protocol included T2-weighted imaging (T2WI), precontrast T1-weighted imaging (T1WI) and dynamic contrast-enhanced (DCE)-T1WI sequences with fat suppression. The scanning parameters and fat suppression technique of each sequence are shown in Table S1.

Clinical and imaging features analysis

The following clinical features were obtained from the electronic medical records of the enrolled patients: age, menopausal status, personal history of breast tumor, personal history of other tumors, family history of breast tumors, family history of other tumors, and location of the mass.

All MG and MR images were analyzed collaboratively by two radiologists with 2 and 8 years of experience in diagnosing patients via breast imaging, according to the American College of Radiology’s Breast Imaging Reporting and Data System (BI-RADS) guidelines (20). Both radiologists were blinded to the pathological grades and a consensus was reached on the final result. In case of disagreement, another senior radiologist with 26 years of experience was involved in the evaluation and provided the final interpretation. For patients with multiple lesions, only the largest mass was evaluated. The maximum diameter of the mass is measured. The following features are analyzed on MG images: breast composition, shape, margin, density, homogeneity, presence of calcifications within the mass, and presence of other associated findings. The evaluated features on MR images include shape, presence of lobulation, presence of cystic components, internal enhancement characteristics, and type of time-signal intensity curve. The DCE-MR images are imported into the Functool software of GE Advantage Workstation (AW) so the most enhanced region of the mass can be delineated as the region of interest (ROI), carefully avoiding areas of cystic necrosis as much as possible. The TIC for the ROI is plotted. TIC curves are classified into persistent type (type I), plateau type (type II), and washout type (type III).

Model establishment

Multicollinearity tests were conducted on the clinical and imaging features selected through univariate analysis. Features with high collinearity (correlation coefficient >0.8) were excluded, and the remaining features were used for multivariate logistic regression analysis to screen for independent factors and to establish a model.

Statistical analysis

All data were analysed using SPSS (v22.0; IBM Corp., Armonk, NY, USA) and R (version 4.2.1; http://www.r-project.org) software. Continuous data were presented as the mean ± standard deviation and were compared using independent sample t-tests or Mann-Whitney U tests as appropriate. Categorical data were presented as case numbers and percentages and were analyzed using Chi-squared tests or Fisher’s exact tests. A P value <0.05 indicates a statistically significant difference. Features with significant differences in univariate analysis were then included in the multivariate logistic regression analysis to identify independent factors. Independent factors with a P value <0.05 in the multivariate analysis were used to establish a model, and the model’s performance was evaluated using the area under the curve (AUC), calibration curve, and decision curve analysis (DCA).

Results

Patient characteristics

A total of 147 patients with FAs (age range, 21–74 years, average age 43.88±9.54 years) and 138 patients with PTs (age range, 16–87 years, average age 46.83±11.25 years) were included in this study. The patients’ characteristics are shown in Table 1. The average age of the patients with PTs was higher than that of those with FAs (P=0.02). Compared to patients with FAs, patients with PTs were more often accompanied by personal history of breast tumors, family history of breast tumors, and family history of other tumors (all P<0.05). In addition, the long diameter of the PTs was significantly larger than that of the FAs (41.87±23.98 vs. 16.80±8.83, P<0.001).

Comparison of MG and MRI features

The most common breast composition included in this study was heterogeneous density, with a significantly higher incidence in patients with FAs than PTs (P=0.001). On MG, FAs are mostly oval-shaped, well-defined, and show equal density, with a small number showing lobulation. Although most the PTs appeared as oval-shaped masses (93/138, 67.4%), the ratio of irregularly shaped PTs was significantly higher than that of irregularly shaped FAs (P<0.001). Moreover, PTs were more frequently high-density masses with lobulation (P<0.001). In addition, FAs had more associated signs, mainly manifested as calcifications around the mass (P=0.007). There was no significant difference in homogeneity or internal calcification between the two types of tumors (Table 2, Figure 2). On MRI, FAs were mostly oval-shaped masses, while PTs were mainly irregularly shaped masses with lobulation (P<0.001), and the latter more commonly showed cystic components and heterogeneous enhancement (P<0.05 for both). In terms of TIC types, both FAs and PTs were mostly type I; however, there was a larger proportion of type II curves than the former (P<0.001) (Table 2, Figures 3,4).

Figure 2 Mammography of fibroadenoma and phyllodes tumor. (A) LCC view and (B) LMLO view: breast fibroadenoma of the left breast. A round mass with equal density is observed in the inner side of the left breast, without lobulation, with slightly blurred boundaries. The maximum diameter of mass approximately 16 mm. (C) LCC view and (D) LMLO view: benign phyllodes tumor of the left breast. An oval mass with lobulation is seen in the lower-inner part of the left breast, with clear boundaries, and a maximum diameter of approximately 65 mm. LCC, lateral cranio-caudal; LMLO, lateral medio-lateral oblique.

Figure 3 Benign phyllodes tumor. (A) T2WI sequence, an oval mass is seen in the lower-inner part of the left breast, with a heterogeneous T2WI signal. (B,C) Show heterogeneous enhancement of the mass after contrast enhancement. (D) The TIC shows a plateau pattern. The X axis represents the strengthening phase, and the Y axis represents the strengthening rate. The number “1” in the figure has no specific meaning. T2WI, T2-weighted imaging; TIC, time-intensity curve.

Figure 4 Fibroadenoma. (A) T2WI sequence, an oval mass is seen on the inner side of the left breast, with a homogeneous T2WI signal. (B,C) Show obvious homogeneous enhancement of the mass. (D) The TIC shows a persistent type. The X axis represents the strengthening phase, and the Y axis represents the strengthening rate. The number “1” in the figure has no specific meaning. T2WI, T2-weighted imaging; TIC, time-intensity curve.

Model construction and evaluation

The significant univariate features mentioned above were included in the multivariate logistic regression analysis, and the following results were obtained (Table 3). Age [odds ratio (OR): 3.51, 95% confidence interval (CI): 1.62–8.04], maximum diameter of the mass (OR: 7.34, 95% CI: 2.89–20.21), density of the mass on MG images (OR: 3.04, 95% CI: 1.43–6.61), lobulation on MR images (OR: 3.79, 95% CI: 1.64–8.97), and type of TIC curve (OR: 4.87, 95% CI: 2.03–12.4) were independent factors for distinguishing FAs from PTs. A model was established based on these independent factors, and a nomogram was used for visualization (Figure 5A). The AUC was 0.90 (95% CI: 0.86–0.94), and internal validation using bootstrapping showed good generalization (Figure 5B). The calibration curve showed good calibration performance (Figure 5C). The Hosmer-Lemeshow test for model goodness-of-fit yielded a nonsignificant statistical result (P=0.33), indicating that the model did not deviate from a perfect fit. The DCA curve demonstrated that the model had good clinical value in distinguishing between the two types of tumors within most risk threshold values (Figure 5D).

Figure 5 Nomogram based on the predictive model (A), bootstrap ROC curve (B), calibration curve (C) and decision curve for the nomogram (D). The blue line represents the ROC curve obtained through 1,000 Bootstrap resampling iterations, while the orange line denotes the mean ROC curve derived from these Bootstrap samples (B). MG, mammography; MRI, magnetic resonance imaging; ROC, receiver operating characteristic; TIC, time-intensity curve.

Discussion

FAs are the most common benign breast tumors (21), while PTs are rare fibroepithelial tumors of the breast, accounting for less than 1% of all breast tumors (1). Although PTs are relatively rare, it is important to accurately differentiate between FAs and PTs preoperatively due to significant differences in their clinical management. In this study, we included an adequate sample (147 cases of FAs and 138 cases of PTs) and compared the clinical and imaging features of FAs and PTs using two imaging modalities (MG and MRI). The results showed that age, maximum diameter of mass, density on MG images, presence of lobulation on MR images, and TIC were independent factors for distinguishing between FAs and PTs. The multimodal fusion model based on clinical and imaging features demonstrated good performance and clinical benefits.

Some scholars believe that age is one of the factors considered when distinguishing between FAs and PTs. Mimoun and others found through a comparative study of 123 cases of FAs and 94 cases of PTs that fibroepithelial lesions of the breast tend to be PTs in patients aged 40 years or older (5). We obtained consistent results. Our results showed that age was an independent factor influencing the accurate differentiation of FAs from PTs. In this study, we also included family and personal history of tumor for analysis and found that compared to FAs, patients with PTs were more likely to have a personal history of breast tumors, a family history of breast tumors, or a family history of other tumors (all P values were lower than 0.05), but none of these factors were independent risk factors for PTs.

Furthermore, size is an important characteristic for distinguishing between FAs and PTs. The maximum diameter of FAs is usually smaller than that of PTs, and a previous study often divided the two types of tumors based on a diameter of 30 mm, categorizing FAs as smaller than 30 mm and PTs as larger than 30 mm (5,22). In our study, the mean maximum diameter of FAs was 16.80±8.83 mm, while that of PTs was 41.87±23.98 mm, which was significantly larger than the former (P<0.001). Multivariate logistic regression results also showed that diameter was an independent factor.

On MG, most patients in our group had a breast composition of heterogeneous density, with a higher proportion of FAs having heterogeneous density. We speculate that this result may be related to the younger age of patients with FAs. PTs were mostly shown as high-density masses (100/138, 72.5%), while FAs were predominantly equal-density masses (104/147, 70.7%), which is consistent with previous reports (19,22).

On MRI, FAs mostly presented as oval masses, while PTs appeared to be more irregularly shaped; there was a significantly higher proportion of PTs showing lobulation than FAs (P<0.001), which was consistent with a previous study (18). Our study results also found that the presence of lobulation on MR images was an independent factor for distinguishing between FAs and PTs. In addition, some scholars believe that the presence of cystic components within the masses is a key factor for distinguishing between FAs and PTs (18,23,24). These differences still existed in our study, although the presence of cystic components within the masses was not an independent factor. Additionally, some researchers have found that the type of TIC is also an identifying factor (18). These studies generally believe that FAs are predominantly characterized by type I curves, while PTs often show type I or type II curves, with type II being more common. In our study, FAs also mostly showed type I curves (130/147, 88.4%), while 66.7% (92/138) and 31.2% (43/138) of the PTs showed type I and type II curves, respectively, and the difference was significant.

In addition to conventional imaging features, previous studies have shown that multiple parameters derived from diffusion-weighted imaging (DWI) of MRI, such as apparent diffusion coefficient (ADC) values, can also be used to differentiate PTs from FAs (25,26). In recent years, with the rapid advancement of artificial intelligence, radiomics features and models based on MG and MRI have also demonstrated promising value in distinguishing between these two types of lesions (27,28). In the present study, we developed a clinical-MG-MRI nomogram as a visualization tool with 5 factors, including age, maximum tumor diameter, tumor density on MG images, lobulation on MR images and TIC. The AUC of the fusion model reached 0.9, indicating good clinical efficacy.

There are several limitations in this study. First, our study is a retrospective and single-center study, so it is necessary to validate the accuracy of the research results through multicenter studies. Second, ultrasound and digital breast tomosynthesis are also used as imaging modalities for FAs and PTs. Due to the rare nature of PTs, only a small number of patients underwent multiple examinations simultaneously. If multiple imaging modalities are included in future studies, a model that can assess the independent factors influencing the accurate differentiation of FAs from PTs will have greater credibility. Finally, this study only analyzed the morphological features of the two types of masses and did not explore the invisible radiomics features inside the masses. Further relevant studies can be conducted in the future.

Conclusions

Breast MG and MRI findings help differentiate between breast FAs and breast PTs preoperatively. Differentiation using the multimodal fusion model is clinically efficacious and beneficial, and thus is potentially useful for accurate clinical diagnosis and treatment.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-145/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-145/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-145/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-145/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study is a retrospective analysis that solely utilized fully anonymized data retrieved from the hospital information system. The Institutional Review Board of Fudan University Shanghai Cancer Center confirmed exemption from ethical approval and waived the requirement for informed consent.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. WHO Classification of Tumours Editorial Board. Breast tumors. Lyon (France): International Agency for Research on Cancer; 2019.
2. Amin AL, Purdy AC, Mattingly JD, et al. Benign breast disease. Surg Clin North Am 2013;93:299-308. [Crossref] [PubMed]
3. Ware AD, Argani P, Cimino-Mathews A. Accuracy and clinical implications of pre-operative breast core needle biopsy diagnoses of fibroepithelial neoplasms and sarcomatoid carcinomas. Breast Cancer Res Treat 2019;178:51-6. [Crossref] [PubMed]
4. Dialani V, Chansakul T, Lai KC, et al. Enlarging biopsy-proven fibroadenoma: Is surgical excision necessary? Clin Imaging 2019;57:35-9. [Crossref] [PubMed]
5. Mimoun C, Zeller A, Seror J, et al. A Pre-operative Score to Discriminate Fibroepithelial Lesions of the Breast: Phyllode Tumor or Fibroadenoma? Anticancer Res 2020;40:1095-100. [Crossref] [PubMed]
6. Tan BY, Acs G, Apple SK, et al. Phyllodes tumours of the breast: a consensus review. Histopathology 2016;68:5-21. [Crossref] [PubMed]
7. Li JJX, Tse GM. Core needle biopsy diagnosis of fibroepithelial lesions of the breast: a diagnostic challenge. Pathology 2020;52:627-34. [Crossref] [PubMed]
8. Tan PH. Fibroepithelial lesions revisited: implications for diagnosis and management. Mod Pathol 2021;34:15-37. [Crossref] [PubMed]
9. Jacobs TW. Update on Fibroepithelial Lesions of the Breast. Surg Pathol Clin 2009;2:301-17. [Crossref] [PubMed]
10. Al-Arnawoot B, Scaranelo A, Fleming R, et al. Cellular fibroepithelial lesions diagnosed on core needle biopsy: Is there any role of clinical-sonography features helping to differentiate fibroadenomas and phyllodes tumor? J Surg Oncol 2020;122:382-7. [Crossref] [PubMed]
11. Li Q, Liu Y, Bai J, et al. miR-140-3p is a potential differential biomarker in benign phyllodes tumors and fibroadenoma of the breast. BMC Womens Health 2022;22:31. [Crossref] [PubMed]
12. Loke BN, Md Nasir ND, Thike AA, et al. Genetics and genomics of breast fibroadenomas. J Clin Pathol 2018;71:381-7. [Crossref] [PubMed]
13. Tummidi S, Kothari K, Agnihotri M, et al. Fibroadenoma versus phyllodes tumor: a vexing problem revisited! BMC Cancer 2020;20:648. [Crossref] [PubMed]
14. Ng CC, Tan J, Ong CK, et al. MED12 is frequently mutated in breast phyllodes tumours: a study of 112 cases. J Clin Pathol 2015;68:685-91. [Crossref] [PubMed]
15. Tan J, Ong CK, Lim WK, et al. Genomic landscapes of breast fibroepithelial tumors. Nat Genet 2015;47:1341-5. [Crossref] [PubMed]
16. Krings G, Chen YY. Genomic profiling of metaplastic breast carcinomas reveals genetic heterogeneity and relationship to ductal carcinoma. Mod Pathol 2018;31:1661-74. [Crossref] [PubMed]
17. Silverman JS, Tamsen A. Mammary fibroadenoma and some phyllodes tumour stroma are composed of CD34+ fibroblasts and factor XIIIa+ dendrophages. Histopathology 1996;29:411-9. [Crossref] [PubMed]
18. Shi Y, Liu Y, Xie W, et al. The useful signs to differentiate breast phyllodes tumor from fibroadenoma using MRI. J Med Imaging 2015;25:1797-801.
19. Wang S, Xu W, Qin G, et al. Differential diagnosis of phyllodes tumor and fibroadenoma of breast using mammography and ultrasonography. Chin J Med Imaging Technol 2019;35:362-6.
20. Sickles EA, D’Orsi CJ, Bassett LW, et al. ACR BI-RADS Atlas, Breast Imaging Reporting and Data System. Reston, VA: American College of Radiology; 2013.
21. Klein S. Evaluation of palpable breast masses. Am Fam Physician 2005;71:1731-8.
22. Yilmaz E, Sal S, Lebe B. Differentiation of phyllodes tumors versus fibroadenomas. Acta Radiol 2002;43:34-9. [Crossref] [PubMed]
23. Kamitani T, Matsuo Y, Yabuuchi H, et al. Differentiation between benign phyllodes tumors and fibroadenomas of the breast on MR imaging. Eur J Radiol 2014;83:1344-9. [Crossref] [PubMed]
24. Mai H, Mao Y, Dong T, et al. The Utility of Texture Analysis Based on Breast Magnetic Resonance Imaging in Differentiating Phyllodes Tumors From Fibroadenomas. Front Oncol 2019;9:1021. [Crossref] [PubMed]
25. Li X, Chai W, Sun K, et al. The value of whole-tumor histogram and texture analysis based on apparent diffusion coefficient (ADC) maps for the discrimination of breast fibroepithelial lesions: corresponds to clinical management decisions. Jpn J Radiol 2022;40:1263-71. [Crossref] [PubMed]
26. Fang J, Zhang Y, Li R, et al. The utility of diffusion-weighted imaging for differentiation of phyllodes tumor from fibroadenoma and breast cancer. Front Oncol 2023;13:938189. [Crossref] [PubMed]
27. Tsuchiya M, Masui T, Terauchi K, et al. MRI-based radiomics analysis for differentiating phyllodes tumors of the breast from fibroadenomas. Eur Radiol 2022;32:4090-100. [Crossref] [PubMed]
28. Zeng F, Zeng H, Yang J, et al. Differentiation Between Phyllodes Tumor and Fibroadenoma of the Breast: A Radiomics Prediction Model Based on Full-Field Digital Mammography & Digital Tomosynthesis. Technol Cancer Res Treat 2024;23:15330338241289474. [Crossref] [PubMed]