Introduction

Although conventional dynamic contrast-enhanced (DCE-) breast MRI, which involves a high spatial resolution and low temporal resolution acquisition, offers high sensitivity for lesion detection, its intermediate specificity results in unnecessary biopsies and limits its use for breast cancer screening. Ultrafast DCE-MRI acquires T₁-weighted images at a high temporal resolution immediately after initial injection of contrast. This enables quantitative assessment of contrast uptake, providing measures of lesion vascular characteristics that could potentially supplement or replace conventional kinetic enhancement features. In this study, we investigated whether quantitative assessment of ultrafast DCE-MRI, integrated into a routine clinical breast MRI protocol, could increase the specificity of breast MRI for lesion diagnosis.

Methods

Subjects and MRI acquisition:
Women who underwent a breast MRI at our institution (March 2019 to August 2020) with a BI-RADS 3, 4, or 5 lesion and subsequent diagnostic biopsy were retrospectively identified for this study. Patients with known breast cancer (i.e., BI-RADS category 6 lesion) were excluded. Imaging was performed on a 3T clinical scanner (Achieva, Philips Healthcare) with a dedicated 16-channel breast coil. Breast MRI exams included two DCE-MRI protocols: 1) a conventional (high spatial resolution) DCE-MRI and 2) an ultrafast DCE-MRI series acquired over the first ~45s after contrast injection (0.1mmol/kg-body-weight delivered at 2cc/sec followed by a 15cc saline flush) (Figure 1). Conventional DCE-MRI, which was only used for lesion localization, was acquired using a fat suppressed, 3D fast gradient echo sequence with one pre-contrast and at least three post-contrast acquisitions centered at 2, 4, and 6 minutes after contrast injection. Ultrafast DCE-MRI was performed during a previously idle delay time using a fat suppressed, 3D fast gradient echo sequence with compressed SENSE (reduction factor=7), TR=2.7ms, TE=1.4ms, α=10º, FOV=240×360cm, matrix=132×199, slice thickness=3mm, with a volume collected every ~8s, and a total of five post-contrast volumes.

Quantitative analysis of ultrafast DCE-MRI:
Prior to quantitative analysis, the ultrafast series was corrected for any motion using a 2-D affine registration (MATLAB, MathWorks, Natick, MA). Subtraction images (post-contrast minus pre-contrast) were created at each ultrafast timepoint. Lesions were qualitatively labeled as “yes” or “no” visible enhancement on subtraction images during the ultrafast series. Lesions with visible enhancement were included in subsequent quantitative analysis, and lesion regions-of-interest (ROIs) were segmented using fuzzy c-means clustering. Both semi-quantitative¹ and quantitative² methods were used to characterize lesion contrast kinetics on ultrafast DCE-MRI. For semi‑quantitative analysis, relative enhancement (%) curves were fit to a half-logistic function yielding voxel-wise measures of contrast bolus arrival time (BAT) that were then normalized to the aorta BAT for each patient. The slope of the relative enhancement curve was used to calculate the maximum slope (MS, %·s ^-1) for each voxel. For quantitative analysis, ultrafast DCE-MRI was fit to a modified Kety-Tofts model to extract pharmacokinetic parameters (K^trans (min^-1) and v_p), using a population arterial input function and assuming a native T₁ value of 1400 ms³. For each measure (BAT, MS, K^trans, v_p), the median was calculated across all voxels within the ROI for each lesion.
Differences between benign and malignant lesions for each imaging measure were calculated using Wilcoxon rank sum test (p<0.05 considered significant). Diagnostic performance was assessed by area under the receiver operating characteristic curve (AUC).

Results

Sixty-nine women were retrospectively identified and evaluated for this study (median age: 47, range: 27-73 years). Of the 72 total lesions, 19 (26%) were diagnosed as malignant and 53 (74%) as benign. Ten lesions exhibited no visible enhancement over the 45s ultrafast timeframe and were all found to be benign. The remaining 62 lesions (malignant: N=19, benign: N=43) were visible on subtraction images and were segmented for quantitative analysis. Malignant lesions demonstrated significantly higher v_p (0.033 vs. 0.023, p=0.03, AUC = 0.67) compared to benign lesions (Figure 2), as well as higher MS, trending towards significance (11.3 vs. 8.0 %·s^-1 p=0.07, AUC = 0.65). No significant differences were observed in K^trans or BAT. A multivariable model including v_p and qualitative assessment of enhancement during the ultrafast series (i.e., “yes” or “no” visible enhancement) showed an improved AUC = 0.73, compared to qualitative assessment alone (AUC = 0.59, Figure 3).

Discussion & Conclusion

These results demonstrate that quantitative assessment of ultrafast DCE-MRI can aid in discrimination of benign and malignant lesions without increasing image acquisition time. It is important to note this ultrafast approach is compatible with both conventional and abbreviated MRI protocols. Malignant lesions were observed to have higher v_p and MS, indicating increased lesion vascularity, supporting findings in previous studies^1,2,4. Additionally, our results suggest absence of enhancement during an ultrafast series (≤45s post-contrast) may help exclude malignancy and reduce unnecessary biopsies. Furthermore, we demonstrated the feasibility of quantitative DCE-MRI in a clinical breast protocol through the estimation of pharmacokinetic parameters from ultrafast data. This quantitative analysis can facilitate comparisons of parameter values across institutions and provide insight into the underlying physiological characteristics of a lesion. Altogether, these findings support the integration of an ultrafast series in clinical protocols to improve specificity of breast MRI and potentially allow for reduced scan times.

Acknowledgements

Supported by NIH grants R01CA203883 and R01CA248192.