Introduction Breast cancer is a prognostically heterogeneous disease including different subtypes with different clinical outcomes and treatment responses [1-3]. Tumor size, axillary lymph node status, histologic grade, estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki-67 labeling index are known prognostic factors of breast cancer [4]. Early detection of prognostic factors may help in predicting clinical outcomes and may also play a significant role in identifying patients who are most likely to benefit from treatment.Synthetic MRI technique is a novel imaging method that allows simultaneous acquisition of multiple relaxometry, including the longitudinal relaxation rate, transverse relaxation rate and the proton density (PD) [5, 6][pywu1] , in a single scan. It has been extensively utilized in imaging of the brain [7, 8], osteoarticular system [9, 10], and prostate [11]. Two studies involving breast cancer had reported the feasibility of synthetic MRI in breast imaging [12, 13]. However, few studies reported the clinical value of synthetic MRI in evaluating prognostic factors and molecular subtypes of breast carcinoma.The purpose of this study is to quantitatively evaluate intra-tumoral heterogeneity through a histogram analysis of quantitative parameters obtained from synthetic MRI, and determine correlations of these histogram characteristics with prognostic factors and molecular subtypes of invasive ductal carcinoma (IDC).
Materials and methods In total, 122 female patients with primary breast IDC were enrolled in the study. Breast MRI was performed on a 3.0T scanner (SIGNA Pioneer, GE Healthcare, Milwaukee, WI). Routine T1WI, T2WI and DWI were first obtained, followed by synthetic MRI and dynamic contrast-enhanced (DCE)-MRI acquisitions. The detailed acquisition parameters were as follows: (1) synthetic MRI: TR/TE1/TE2 = 4243/17/93 ms, FOV = 360 × 360 mm, matrix = 256 × 256, slice thickness = 5.5 mm; (2) DCE-MRI: TR/TE = 7.2/2.4 ms, flip angle = 10, FOV = 350 × 329 mm. Five post-contrast scans were acquired after intravenous injection of a contrast agent at 2 ml/s with a dose of 0.2 mmol/kg. Data of synthetic MRI was processed using SyMRI 8.0 (SyntheticMR, Linköping, Sweden) to generate quantitative parametric maps (T1, T2, and PD). We first performed a rigid registration between images of CE-MRI and processed MAGIC images. Then ROI of the whole tumor was manually delineated on registered CE-MRI images using ITK-SNAP (http://www.itksnap.org/pmwiki/pmwiki.php). Finally, the ROI was applied to extract histogram features, including mean, minimum, maximum, 10th and 90th percentiles, kurtosis, and skewness, using PyRadiomics [14][pywu2] . Histopathologic results were analyzed by two pathologists. Molecular subtypes were defined based on the St. Gallen consensus, as follows: luminal A (positive for ER and/or PR, with Ki-67<14%); luminal B (positive for ER and/or PR, with overexpressed HER2 or Ki-67≥14%); HER2 (negative for ER and PR, with overexpressed HER2); TN (negative for ER, PR, and HER2).Statistical analyses were performed using SPSS 19.0 (IBM, Armonk, NY). The histogram characteristics of quantitative parameters were compared between types in different prognostic factors using Mann-Whitney U test and among molecular subtypes (luminal A, luminal B, and TN) using Kruskal-Wallis test, followed by the post-hoc test. P value < 0.05/7 = 0.0071 (control for multiple comparisons across seven prognostic factors) for the comparison of prognostic factors and P value < 0.05 for the comparison of subtypes, were considered statistically significant.
Results For comparisons of histogram characteristics between types in different prognostic factors, we found that PDmedian was significantly higher in cancers with HER2 positivity than that with HER2 negativity. Cancers with high nuclear grade had significantly higher PD10th, PDmean, PDmedian, PDmax and PDkurtosis. The T110th was significantly higher in cancers with PR negativity than that with PR positivity. Cancers with hormone receptor negativity had significantly higher T210th, T2mean and T2median than that with hormone receptor positivity. Cancers with high nuclear grade had significantly higher T210th, T2mean, T2median and T2max. T210th, T2mean, T2median and T290th were found significantly different among tumor subtypes. The post-hoc test revealed that T210th, T2mean, T2median and T290th in TN were significantly higher than luminal A subtypes. Figure 1 shows the ROC results of T2 histogram characteristics in distinguishing TN from luminal A subtypes. T210th, T2mean, T2median and T290th in TN were significantly higher than luminal B subtypes. Figure 2 shows the ROC results of T2 histogram characteristics in distinguishing TN from luminal B subtypes. T2meanand the T2median values of TN, luminal B and luminal A types were arranged in descending order (Figure 3).No significant differences were found in PD and T1 values among tumor subtypes.
Discussion Breast cancer heterogeneity influences poor prognosis thorough therapy resistance. Here we performed a histogram analysis of T1, T2, and PD quantitative parameters obtained from synthetic MRI. We specifically investigated associations of the extracted histogram characteristics with several prognostic factors and molecular subtypes of invasive ductal breast cancers. Overall, we demonstrated that whole-tumor histogram-based imaging features derived from synthetic MRI correlated with prognostic factors and may be considered a non-invasive approach for discriminating IDC subtypes.

Acknowledgements

No acknowledgement found.