MRI is of high importance in breast cancer imaging. Contrast-enhanced (CE) measurements can reveal the tumor perfusion and thus serve to classify and characterize lesions. However, especially in screening situations, repetitive administration of contrast agents should be avoided if possible¹.

Arterial Spin Labeling (ASL) is an established technique for measuring perfusion without contrast agents in the brain², but in other organs, ASL remains challenging. Only few groups have shown the feasibility of Arterial Spin Labeling (ASL) in the breast^3-8 using either 2D single-slice or multi-slice imaging.

In this breast cancer case study, we present a fast 3D single-shot ASL sequence that is designed to provide high SNR and to cover a large volume within reasonable acquisition time.

The patient measurements for this case study have been performed at the Department of Radiology of Radboud University Medical Center, Nijmegen, Netherlands, on a Siemens Magnetom Skyra 3T system (Siemens Healthcare, Erlangen).

The pulsed ASL sequence (PASL) uses a FAIR⁹ labeling with a 3D-GRASE readout¹⁰. In an almost coronal orientation we acquire a volume with a 96x48x8 matrix and a voxel size of 3.6x3.6x4mm³ in single-shot mode. 80 label and control pairs are acquired within 7:36 min. A relatively long inflow time of TI=2000ms has been chosen to measure the tumor enhancement. Perfusion is not quantified but instead, the sequence is used to qualitatively specify lesions. Background suppression with 4 adiabatic inversion pulses¹¹ has been optimized to minimize artifacts related to signal from fat and static tissue.

The slice-selective inversion volume of the FAIR labeling was chosen to be identical with the imaging volume, which means that only blood outside the imaging volume is labeled. Therefore, this volume needs to be positioned very carefully in order to label all inflowing blood (Fig.2). By keeping a small gap between chest wall and breast we ensure that blood is labeled in all feeding vessels.

Each image was acquired in single-shot mode, which makes the sequence robust against motion. Still, motion can also occur between label and control image as well as between different repetitions. Therefore, we manually excluded affected label and control pairs from the evaluation but we did not correct for between-repetition movement as the latter does not affect measured data significantly. Finally, the individual measurements were averaged.

The examined patient has a cystic carcinoma with a diameter of 3 cm in the right breast (Fig.1). The tumor tissue around the cystic part of the tumor is highly perfused as the CE subtraction images show.

12 label and control pairs of the ASL data have been corrupted by motion and removed manually before averaging.

In Fig.3 the CE image has been reformatted to match the ASL slices. For the tumor perfusion, the ASL perfusion-weighted image corresponds well to the CE image. When the ASL perfusion weighted image is reformatted to the CE image (Fig.4), the correspondence becomes even more obvious.

Using our approach it is possible to obtain perfusion images of breast lesions without contrast agent. However, a major drawback of the demonstrated sequence is that the tumor position must be known before scanning as the thickness of the volume does not offer full-breast coverage. A thicker volume could be acquired using multiple segments but also the FAIR labeling would hamper a complete imaging of the breast.

Furthermore, spatial resolution and SNR are inferior to CE sequences and might hide smaller lesions. In addition, vessels perpendicular to the slice orientation also appear bright in the subtraction images and thus might lead to false positive findings.

Finally, by applying this method to more subjects, parameters like the necessary number of measurements and the optimal inflow time (TI) must be optimized further.