Imaging characteristics of young age breast cancer (YABC) focusing on pathologic correlation and disease recurrence

The purpose of this study is to investigate imaging characteristics of young age breast cancer (YABC) focusing on correlation with pathologic factors and association with disease recurrence. From January 2017 to December 2019, patients under 40 years old who were diagnosed as breast cancer were enrolled in this study. Morphologic analysis of tumor and multiple quantitative parameters were obtained from pre-treatment dynamic contrast enhanced breast magnetic resonance imaging (DCE-MRI). Tumor-stroma ratio (TSR), microvessel density (MVD) and endothelial Notch 1 (EC Notch 1) were investigated for correlation with imaging parameters. In addition, recurrence associated factors were assessed using both clinico-pathologic factors and imaging parameters. A total of 53 patients were enrolled. Several imaging parameters derived from apparent diffusion coefficient (ADC) histogram showed negative correlation with TSR; and there was negative correlation between MVD and Ve in perfusion analysis. There were nine cases of recurrences with median interval of 16 months. Triple negative subtype and low CD34 MVD positivity in Notch 1 hotspots showed significant association with tumor recurrence. Texture parameters reflecting tumor sphericity and homogeneity were also associated with disease recurrence. In conclusion, several quantitative MRI parameters can be used as imaging biomarkers for tumor microenvironment and can predict disease recurrence in YABC.


Correlation between pathologic factors and imaging parameters. Apparent diffusion coefficient
histogram and tumor-stroma ratio. Of the 52 patients who performed tumor-stroma ratio (TSR) analysis, 43 patients were stroma-poor type and 9 were stroma-rich type. Mean apparent diffusion coefficient (ADC) value of stroma-poor type was significantly lower than that of stroma-rich type (p = 0.006) ( Table 2). Negative correlation was revealed among TSR and several parameters of ADC histogram: minimum value, mean value, median value, 5th percentile, 10th percentile, 25th percentile, 75th percentile and 90th percentile (Table 3).
Perfusion parameters and endothelial notch 1 and microvessel density. The endothelial Notch 1 (EC notch 1) and microvessel density (MVD) analyses were performed in 45 patients of 53 patients. There was no association between perfusion parameters and EC notch 1. Of the multiple parameters from perfusion analysis, only the mean value of Ve showed negative correlation with MVD (correlation coefficient = −0.30, p < 0.046).  www.nature.com/scientificreports/ Recurrence associated factors. Clinico-pathologic factors. There were 9 cases of tumor recurrence including loco-regional recurrence and distant metastases. The recurrence group showed significantly higher estrogen receptor (ER) negativity and progesterone (PR) negativity (p = 0.02, 0.044) than in non-recurrence group; and Ki-67 index tended to be higher in recurrence group (57.6 ± 32.7 [SD], p = 0.056). Lack of adjuvant hormone therapy was also higher in recurrence group (p = 0.024). The ER negativity, PR negativity and lack of adjuvant hormone therapy also showed significantly increased risk for disease recurrence (Table 4). Ten cases of the total 53 cases were triple negative subtypes. Of these 10 cases, 5 cases had disease recurrence accounting for 55.6%   In immunohistochemical staining, lower positivity of CD34 MVD in Notch1 hot spots was significantly associated with disease recurrence (OR 9.00, 95% CI 1.08-21.90, p = 0.02) (Fig. 2, Table 4).  www.nature.com/scientificreports/ Imaging factors.

Qualitative parameters: morphologic characteristics
The mean size of the tumor was 2.9 cm (mean 2.88 cm ± 2.05 [SD]). In the recurrence group, the mean size of tumor was 3.9 cm (3.9 cm ± 1.7 [SD]), which was significantly larger than that of the tumor in non-recurrence group (mean 2.7 cm ± 2.1 [SD], p = 0.02). However, the increase in the mean size of tumors was not significantly associated with disease recurrence (p = 0.081).
More than half of the mass cases manifested as oval shape (n = 27, 57.5%) with circumscribed margin (n = 26, 55.3%). Except for one case of mucinous carcinoma which lacked enhancement, all mass lesions showed heterogenous or rim enhancement. However, these morphologic features showed no significant difference between recurrence and non-recurrence group.
The peritumoral edema was found only in 14 cases (26.4%). It was more frequently found in the recurrence group (55.6%, p = 0.044), and it was a significant factor for disease recurrence (OR 4.861, 95% CI 1.08-21.90, p = 0.039).
Mass shape, distribution of non-mass enhancement, number of prominent vessels around the index cancer, and other non-tumor factors showed no significant association with disease recurrence. The difference in whole qualitative parameters including morphologic characteristics of tumor and non-tumor factors between recurrence and non-recurrence groups is provided as Supplementary Table 1.

Quantitative parameters
From texture analysis, multiple parameters showed significant difference between recurrence and non-recurrence group (Supplementary Table 2). Among these parameters, low surface area to volume ratio calculated from T2-weighted imaging (T2WI) (OR 15.625, 95% CI 2.94-82.84, p = 0.001) and ADC map (OR 16.949, 95% CI 2.93-99.44, p = 0.002) showed the highest risk increase from other parameters. Of the first order features, only parameters calculated on the ADC map were correlated with disease recurrence; however, features such as entropy, kurtosis or skewness, which are commonly referred to describe tumor texture features, were not included. Of the second-order features, one gray level co-occurrence matrix (GLCM) feature calculated on the ADC map showed increased risk for the disease recurrence: inverse difference moment (IDM) (OR 5.625, 95% CI 1.23-25.76, p = 0.026), which is a measure of local homogeneity. In similar context, lower value of run length non uniformity normalized feature, a kind of homogeneity in image, showed the tendency to increase the risk of disease recurrence (OR 8.00, 95% CI 1.08-21.90, p = 0.059) ( Table 5).
There was no significant association between perfusion parameters, and ADC histogram and recurrence in our study.

Discussion
In this study, imaging characteristics of YABC were assessed focusing on their correlation with pathologic factors and on association with disease recurrence.
In the correlation between pathological factors and imaging parameters, TSR and ADC value, and MVD and Ve value showed correlation with each other. These were consistent with previous studies [13][14][15] . However, these pathologic factors and imaging parameters showed no significant correlation with disease recurrence in our study. The association between TSR and disease recurrence showed results that contradict the previous studies. In one previous study, stroma-rich type tumor showed higher relapse rate in triple negative breast cancer 16 . However, five cases of triple negative subtype with disease recurrence were all stroma-poor tumors in our study.
Although there was no significant association between EC Notch1 and image parameters, CD34 MVD in Notch1 hot spots showed negative correlation with disease recurrence. Despite that higher MVD showed association with poor prognosis of tumor in general, CD34 MVD showed contradictory result to previous studies 17 . This result may be due to the limitation of the small number of cases in our study. However, CD34 MVD in Notch 1 hot spots may be a new prognostic factor with different means from CD34 MVD measured in the entire tumor tissue. Further research is needed in the future for validation of this result.
According to previous studies on imaging characteristics of YABC, breast cancer under 40 years of age showed irregular shape with irregular or spiculated margin and internal heterogenous enhancement pattern on MRI 18,19 . These are typical findings of breast malignancy in average age women. However, more than half of the malignancies in our study manifested as mass lesions with oval shape with circumscribed margin, which is different from previous studies 20 . In addition, low value of surface area to volume ratio from T2WI and ADC map showed the highest correlation with recurrence on texture analysis. In other words, the more spherical the three-dimensional shape of the tumor is, the higher the association with recurrence. These morphologic features are similar to triple negative subtype breast cancer, which generally shows as mass with smooth or circumscribed margin 21,22 . Considering the high recurrence rate of triple negative subtype 23,24 , the results are consistent with previous studies. Moreover, triple negative subtypes accounted for more than half of the recurrence group in our study. Our results are not only consistent with previous studies, but also are more meaningful because they have been more objectively verified through texture analysis.
High value of IDM from ADC map also showed correlation with disease recurrence. This result is not consistent with the results of previous studies because the IDM indicates homogeneity of the tumor. In a previous study, the ADC value difference of tumor could reflect tumor heterogeneity; and high ADC value difference was associated with distant metastasis in breast cancer patients with mean age of 50 years old 25  www.nature.com/scientificreports/ the ADC value difference was measured only on one section with the widest diameter of tumors. In our study, texture parameters of ADC map were measured in three dimensions to reflect the heterogeneity or homogeneity of the ADC value of the entire tumor. Another previous study has proven increased uniformity and decreased entropy in high histologic grade tumor 26 , though the texture parameters were measured on contrast-enhanced T1-weighted subtraction images. These results, which are inconsistent with previous studies, may be the unique features of young age breast cancer because there are few studies on pathologic factors and quantitative MR parameters focusing on breast cancer in women under 40 years old.
Several texture parameters from ADC map showed significant correlation with disease recurrence whereas some previous studies showed that texture analysis of ADC map could not help predict the prognosis of breast cancer 27,28 . Diffusion weighted imaging (DWI) is emerging as a non-contrast imaging to replace contrastenhanced MRI due to safety from gadolinium-based contrast agents. Thus, our results can be regarded as meaningful since many significant results were obtained from DWI. Based on this, DWI may have the potential to be a surveillance tool for high-risk young women who need annual contrast enhanced breast MRI by recommendation of American College of Radiology 29 .
There are few limitations in our study. Despite its prospective nature, this is a single center study; and the number of enrolled patients and event number are small, which could weaken the representatives of characteristics of YABC. In addition, the follow-up period after surgery was not enough, considering disease free survival of hormone receptor positive breast cancer. There were also some missing data in pathological factors and MR parameters.
Despite these limitations, this prospective study would have the novelty for description of imaging characteristics of YABC focusing on the correlation with pathologic factors and disease recurrence. Low CD34 MVD in Notch 1 hot spots and tumor homogeneity on ADC map, which showed correlation with disease recurrence, were contradictory to results of previous studies. These can be the unique features of YABC, although further validation would be needed with larger number of patients and longer follow-up period. In the situation where Asian breast cancer with a high proportion of young age women is rapidly increasing, studies to reveal the characteristics of YABC are necessary. Therefore, we expect this study to be the milestone of future studies that may reveal the characteristics of YABC.

Materials and methods
Patients. This prospective study was approved by institutional review board of Seoul St. Mary's Hospital (reference number, KC16EISI0425) and assessments were carried out as per the rules of the Declaration of Helsinki of 1975, revised in 2013. From August 2017 to August 2019, patients diagnosed with invasive breast cancer under 40 years of age were the subjects for this study. Every patient was provided with informed consent before enrollment and agreed to undergo pre-treatment breast MRI with specific sequences in our institution for this study. DWI was performed before dynamic contrast enhanced MRI, using a single-shot spin-echo EPI pulse sequence with TR/TE 12,043.5/102.3 ms, flip angle of 90°, FOV of 320 × 320 mm 2 , matrix size of 184 × 184, slice thickness of 3 mm, using three b values (b = 0, 1000 s/mm 2 ). The ADC map was calculated with monoexponential fit using a b value of 1000. www.nature.com/scientificreports/ Clinical and pathologic information. Medical records of all enrolled patients were reviewed. Clinical information included patient's age, date of surgery, the last date of outpatient clinic after surgery, date of disease recurrence, and the type of treatment received including surgery, chemotherapy, hormonal therapy and target therapy. Pathologic information of breast cancer included histologic type and grade, hormonal receptor status, human epidermal growth factor receptor 2 (HER2) status, Ki-67 index, and pathologic TNM staging by AJCC 7th. Several additional pathologic analyses were performed including TSR, EC notch 1 and MVD of tumor. These factors were known as prognostic factors in previous studies 16,30,31 and showed some correlation with imaging parameters: ADC values and perfusion parameters. EC Notch 1 has not been identified to correlate with imaging parameters to our knowledge. However, since this factor is already known as an angiogenesis-reflecting factor, we expect it to be able to predict the association with perfusion parameters; Ktrans, Kep and Ve.

MRI.
These analyses were performed with surgical specimen. However, the tissue obtained from the core-needle biopsy at the time of cancer diagnosis was used if the patient underwent neoadjuvant chemotherapy. Staining of MVD and EC Notch 1 was omitted in cases of biopsy-tissue acquired in other hospitals.
Recurrence was defined as one of the followings: newly diagnosed breast cancer, axillary lymph node metastasis, or distant metastasis after surgery. Recurrent diseases were diagnosed by pathological confirmation when biopsy was possible; and in cases where biopsy was unavailable, other imaging modalities such as computed tomography (CT), or positron emission tomography (PET) -CT were used.
Ethical approval. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1975 Helsinki declaration and its later amendments or comparable ethical standards.
Informed consent. Informed consent was obtained from all individual participants included in the study.

Imaging analysis
Morphologic analysis. Two breast expert radiologists prospectively analyzed morphologic characteristics of tumors on breast MRI in consensus, blind to the clinico-pathologic information of patients. The longest diameter of tumor, shape, margin, internal enhancement pattern, and kinetic curve were analyzed by BI-RADS lexicon on early and late dynamic phase of DCE-MRI. Presence of peritumoral edema and number of prominent vessels around the tumor, which were known as imaging prognostic factors by previous studies, were assessed on T2WI and on maximal intensity projection (MIP) image, respectively [32][33][34] . Fibroglandular tissue (FGT) pattern on T2WI, and background parenchymal enhancement (BPE) pattern on MIP images were also analyzed as nontumor features in the contralateral breast, which was not involved by cancer.
Quantitative analysis of MR parameters. Apparent diffusion coefficient histogram. Post-processing was performed using In-house analysis software (EXPRESS; licensed by Jinwoo Hwang, Philips Healthcare Korea) to obtain ADC histogram of a defined tumor volume. Freehand type of region of interest (ROI) was manually drawn along the margin of the lesion for each image of ADC data. Then, a volume of interest (VOI) was defined as a stack of ROIs; and minimum, maximum, mean, 5th, 10th, 25th, 75th, 90th, 95th percentile ADC values, skewness and kurtosis of the pixels contained within the VOI were calculated (Fig. 3).
Perfusion parameter and texture analysis. The analysis for perfusion parameters was conducted in the same way as the previous study from our institution using a post-processing software (Olea Sphere, version 3.0, licensed by Olea Medical, La Ciotat, France) 20 .
Texture analysis was also performed in Olea Sphere, with the same VOI drawn for perfusion parameter. We extracted first order features, gray level co-occurrence matrix (GLCM), and gray level run length matrix (GLRLM) from calculation on subtracted early dynamic phase T1WI, T2WI and ADC map. If the tumor was not clearly visualized on T2WI or ADC map, the VOI drawn on subtracted early dynamic phase T1WI was automatically applied at the similar location on T2WI or ADC map in the program and texture parameters were calculated (Fig. 4).

Pathologic analysis and immunohistochemistry.
Tumor-stroma ratio. One HE-stained slide of representative tumor was analyzed on microscopic examination. As previously determined, cases with stroma proportion of < 50% were considered as stroma poor and cases with stroma proportion of ≥ 50% were regarded as stroma rich (17).
Angiogenesis assessment with CD34 and Notch1 immunohistochemical staining. Four µm-thick sections for immunohistochemical staining of CD34 and Notch1 were obtained from formalin-fixed paraffin embedded tissue. The used primary antibodies were anti-CD34 mouse monoclonal antibody (clone QBEnd 10, 1:100 dilution, Dako, Glostrup, Denmark) and anti-Notch1 rabbit monoclonal antibody (clone D1E11, 1:50 dilution, CST, Danvers, USA). The CD34 staining was performed using the DAKO Omnis automated immunohistochemistry www.nature.com/scientificreports/ system (Dako). Tissue sections were deparaffinized with xylene, three times for 10 min, for Notch1 staining. Then, the sections were rehydrated using 100%, 95%, and 70% graded ethanol for 5 min, each after oven incubation at 60 °C for 1 h. Antigen retrieval was carried out in a pressure cooker (Electric Pressure Cooker CPC-600, Cuisinart, East Windsor, USA) for 20 min using 1× citrate buffer (pH 6.0). The endogenous peroxide activity was blocked by methanol-diluted 3% hydrogen peroxide for 15 min. Sections were incubated with the primary antibody for 1 h at room temperature (22-25 °C) in a humidified chamber. The immunoreaction signal was amplified and was revealed using the Polink-2 HRP DAB Broad-2 Detection system kit (GBI, Mukilteo, USA). Subsequently, these sections were counterstained with Harris's hematoxylin (YD Diagnostics, Yongin, Korea). We determined the MVD with CD34 immunostained tumor section. We counted the microvessels in the highest CD34 positive neovascularization areas (hot spots) of the tumor stroma. These were identified on the scan view in accordance with the method described by Weidner N (18). In the Notch1 immunostained slides, the highest Notch1 positive neovascularization areas (hot spots) were selected and counted Notch1 positive microvessels (Notch1 MVD). In the CD34 immunostained slides, the corresponding areas to Notch1 hot spots were selected and counted CD34 positive microvessels (CD34 MVD). The ratio of Notch1 MVD to CD34 MVD was defined as endothelial-Notch1 as previously described 31 . The EC Notch1 was classified into EC Notch1 low group or EC Notch1 high group based on the cutoff value. The cutoff value of EC Notch1 was set at the median expression value.
Statistical analysis. For analysis of characteristics of young age breast cancer, independent t test or Mann-Whitney test was used for continuous variables, and Chi-square or Fisher's exact test was used for categorical variables. The correlation between MR parameters and pathologic factors was analyzed using Pearson or spearman correlation coefficient. The association between recurrence and baseline and clinical characteristics were performed to logistic regression analysis. All statistical analyses were used by SAS Version 9.4 (SAS Institute Inc, Cary, NC) and p value below 0.05 was considered statistically significant.  The color maps that do not match the tumor were manually modified using the original MR images of these sequences.