The prospective study protocol was approved by our institution’s ethics committee. Between November 2020 and February 2021, patients with histologically proven invasive breast carcinoma, a tumor size larger than 15 mm on ultrasound, and who were referred for pretreatment evaluation, underwent breast MRI. Thirty consecutive patients were invited in this study and written informed consent was obtained from all participants.
Mr Image Acquisition
Studies were performed on a 1.5 T scanner (Ingenia, Philips Healthcare, Best, The Netherlands) using a 16-channel breast-phased array coil. Images were acquired in the prone position and in axial orientation to cover the whole breasts. Participating patients underwent synthetic MRI in addition to routine clinical MRI sequences. In synthetic MRI, quantification of T1 and T2 relaxation rates as well as PD was performed using the QRAPMASTER pulse method sequence, which is a multislice, multiecho, and multisaturation delay acquisition sequence . Two sets of echo times (TEs) and 4 sets of delay times were used to generate 8 complex images in each section in order to quantify T1, T2, and PD. The TEs were 12.5 and 100 msec, the delay times were 151, 604, 2115, and 4382 msec, and the repetition time (TR) was 4462 msec. Thirty slices were acquired, voxel size was 1.42 × 1.92 × 4.00 mm3, and total acquisition time was 3:40 minutes. Synthetic images were created using SyMRI StandAlone software (SyntheticMR AB, Linköping, Sweden). Following synthetic MRI, routine T1-weighted images (TR/TE, 600/12 msec), STIR (TR/TE/TI, 3500/90/160 msec), diffusion-weighted images (DWI) using single-shot echo planar imaging (b values of 0 and 1500 s/mm2; number of excitations, 1 and 14; TR/TE/TI, 6407/86/180 msec; field of view (FOV), 340 mm; matrix, 112×112; section thickness, 4 mm), and three-dimensional (3D) fat-suppressed dynamic contrast-enhanced MRI (TR/TE, 4.0/2.1 msec; flip angle, 15°; FOV, 340 mm; matrix, 352 × 352; slice thickness, 1 mm; intersection gap, 0 mm) were obtained. Dynamic scans were performed before and 4 times after the rapid injection of a bolus of 0.1 mmol/L of gadolinium-based contrast medium per kilogram of body weight.
MR images were evaluated by a board-certified radiologist with 20 years of experience in breast MRI. Dedicated analysis software for synthetic MRI automatically calculated a value of R1 (i.e. 1/T1), R2 (i.e. 1/T2), and PD for all pixels in the ROI. The PD level of pure water was set at 100%. Synthetic T1-weighted images (TR/TE, 600/10 msec) were made for measurement purposes. For measurement of T1, T2, and PD, the radiologist selected the slice that showed the largest slice of the tumor on contrast-enhanced images. Then the corresponding slice was selected on synthetic T1-weighted images. A freehand region of interest (ROI) was placed on the synthetic T1-weighted image while cross-referencing to the contrast-enhanced images, and with the radiologist being blinded to the clinical and pathologic information. A freehand ROI was placed just within the inner border of the tumor, in order to avoid the inclusion of surrounding fat. Mean R1, mean R2, and mean PD shown on the viewers were recorded. Mean T1 and mean T2 were calculated based on the following equation: T1 = 1/R1, T2 = 1/R2. Dedicated analysis software automatically calculated apparent diffusion coefficients (ADC) maps. The same radiologist measured the mean ADC of breast cancers in a similar manner with freehand ROIs set on diffusion-weighted images. The ROIs were copied to the ADC maps to obtain the ADC value.
All pathological specimens were evaluated by our institution’s pathology department using their standard clinical protocol. Pathologists evaluated the specimens on the basis of the World Health Organization histologic classification of breast tumors . The expressions of ER and PgR were assessed by immunohistochemical staining. Per our institutional protocol, any expression of ER or PgR greater than 1% was considered positive [18, 19]. Ki-67 immunoreactivity was evaluated by using the percentage of immunoreactive tumor cells . Based on Ki-67, cancers were divided into low-proliferation group (Luminal A) (Ki-67 < 14) and the high-proliferation group (Luminal B) (Ki-67 ≧ 14) [3, 4].
Data were analyzed using MedCalc 19.8 (MedCalc, Ostend, Belgium). All data are expressed as means ± standard deviation, unless otherwise specified. Normal distribution of the T1, T2, PD, and ADC values was tested using the Shapiro–Wilk test. Age, tumor diameter, Ki-67, and the quantitative MR parameters regarding ER status (positive vs. negative) and Luminal types (A vs. B) were compared using the unpaired t test (for normally distributed data) or the Mann-Whitney U Test (for not normally distributed data). Quantitative MR parameters m including ADC, were correlated with Ki-67 using a Pearson’s correlation analysis (for normally distributed data) or a Spearman correlation analysis (for not normally distributed data). Among resected tumors, the correlation of quantitative MR values with Ki-67 of both biopsy specimen and surgical specimens was assessed.
The effectiveness of quantitative MR parameters, including ADC, in differentiating the ER-negative group from the ER-positive group and Luminal B from the Luminal A was evaluated using receiver operating characteristic (ROC) analysis. For sensitivity and specificity, the optimal cut-off point was determined using Youden index. The area under the ROC curve was expressed as a mean and 95% confidence interval. All statistical tests were two-tailed and significance was set at a P < .05.