Magnetization transfer (MT) imaging is sensitive to changes in the macromolecular content of tissue and is, therefore, gaining increased attention as a noninvasive approach to probe the complex tumor environment in cancer^1,2. We previously demonstrated the reproducibility of magnetization transfer ratio (MTR) measurements in fibroglandular tissue of the breast³; however, MTR is a semi-quantitative measurement dependent upon several factors, including the pulse sequence parameters and tissue relaxation characteristics. Quantitative MT (qMT) imaging is a technique that potentially provides more specific information on tissue composition, including the ratio of macromolecular protons to the protons in the free water pool, or pool size ratio (PSR)⁴. The aim of this study is to explore the reproducibility of PSR measurements of fibroglandular (FG) tissue in healthy controls at 3T.Six women (mean age: 40 years, range: 27-54 years) with no history of breast disease were scanned twice during a single visit: the protocol was run once, the volunteers were taken off the table to stretch then repositioned, and the protocol was repeated. Images were acquired with a 3T Achieva MR scanner equipped with the MammoTrak table, including a dedicated 16-channel sensitivity encoding (SENSE) receive double-breast coil (Philips Healthcare, Best, The Netherlands). For qMT imaging, a MT-prepared (20 ms sinc-Gauss pulse), segmented EPI sequence (5 lines/shot) with a water-selective excitation pulse (1331, 6°), flow-compensation, and respiratory gating was employed. Data were collected at four MT offset frequencies (1, 2, 4 and 8 kHz) using two MT pulse angles (500° and 800°, plus one image at 0° for normalization), resulting in 9 volumes in a nominal scan time of 1 min 38 s. Additional parameters included: acquired (reconstructed) resolution = 2×2×5 mm³ (1.33×1.33×5 mm³), TR/TE = 48/6.6 ms, and SENSE factor = 1.5. The qMT model requires independent T₁, RF transmit (B₁⁺), and main magnetic field (∆B₀) estimates. T₁ was estimated using the multiple flip angle (MFA) method with 10 flip angles (2:2:20°) and TR/TE = 7.9/4.6 ms. B₁⁺ was measured using a Bloch-Siegert method⁵. ∆B₀ was measured using a dual-gradient echo method with fat and water protons in phase. All data were non-rigidly co-registered to the reference qMT volume (Advanced Normalization Tools) and fit to a two-pool model to estimate PSR and the T₂ of the macromolecular protons. During fitting, the T₁/T₂ of water protons (40), and the MT rate (48 Hz) were fixed to published values in skeletal muscle⁶. The reproducibility of the median PSR (mPSR) for fibroglandular (FG) tissue was determined by evaluating the 95% CI of the mean difference, within-subject standard deviation (wSD), and repeatability coefficient (r) with the test-retest sessions⁷.Median fibroglandular PSR values for each volunteer are listed in the Table. PSR maps from the central slice of each subject for scan 1 (top row) and scan 2 (bottom row) are shown in Figure 1. FG PSR values are plotted for each data set, with the median values denoted by the solid lines (Figure 2a). The difference between the mPSR values between scans is plotted against the average of the mPSR values for the two scans (Figure 2b). The mean difference for all subjects (0.15) was not significantly different from zero, and the individual difference values were not dependent upon the average mPSR value. The 95% confidence interval limits were ±1.07 (α = 0.05) and the repeatability measure (2.77 × wSD) was 2.07.

qMT imaging is potentially sensitive to changes in glycoprotein content in the extracellular matrix, which is known to change during the tumor life cycle and in response to treatment⁸; thus, it is a reasonable hypothesis that qMT parameters may change during the course of therapy and potentially be useful for predicting treatment response. However, prior to applying qMT imaging to assess treatment response, it is necessary to have an expectation of the typical variation in the PSR of healthy breast tissue. To the authors’ knowledge, this is the first application of qMT in the breast, resulting in reasonable reproducibility of PSR measurements with a clinically feasible sequence.

It is noted that the accuracy of the model fit may be reduced because of several factors, including a reduced number of offsets to keep the qMT sequence as short as possible, the fixed values used in the fitting process, partial volume averaging of adipose and fibroglandular tissue in the relatively large voxels, and volunteer motion between the B₁⁺, B₀, T₁, and qMT acquisitions. Future work includes addressing these issues and applying this technique in an ongoing multi-parametric study to predict treatment response⁹.