CEST, specifically APT neuroimaging, is emerging as a powerful tool in assessment and characterization of tumor aggressiveness and treatment response^1,2. Recent studies have demonstrated application of CEST to breast malignancies and several characteristic frequencies were proposed^3-5. One of the difficulties of CEST-MRI in breast is the large fat content, leading to lipid artifacts and potentially degradation of the CEST effect.

We hypothesize that CEST-MRI can provide metabolic information⁶ to aid tumor characterization for enhanced specificity and predictive value of MR mammography. To remove lipid artifacts, the previously proposed CEST-mDixon method^7,8 is used to obtain pure water CEST images. The mDixon technique is insensitive to B₀ and B₁ inhomogeneity and does not interfere with the saturation pulse train. Moreover, the B₀ maps derived from the mDixon technique can be used for field inhomogeneity correction, without the need for a separate B₀ mapping. The purpose of this work is to explore the potential of mDixon-based CEST-MRI for breast lesions characterization and identify the CEST frequency ranges with the highest correlation with malignancy, without fat contamination.

Four female volunteers, 1 healthy and 3 breast cancer patients prior to biopsy, were scanned on a 3T MRI (Ingenia, Philips Healthcare) using a 16-channel breast coil. CEST images were acquired using a 2D multi-shot T1-weighted turbo field echo (TFE) sequence with 3-point multi-echo Dixon with TR/TE₁/ΔTE = 5.1/1.57/1.0 ms. We have chosen 3 TE values, since this is the least number of echoes needed to robustly separate water, fat, and B₀. ΔTE was adjusted to the minimum possible value thus reducing potential artifacts by phase-wrapping. The imaging slice was placed for optimal observation of the fibroglandular tissue and/or the tumor. The CEST saturation pulse train consisted of 10 hyperbolic secant pulses, 49.5 ms each, with inter-pulse delay 0.5 ms and FA=900^o (B_1rms=1.2 μT). 33 points in the Z-spectrum from -6 ppm to 6 ppm were acquired. Other imaging parameters included centric ordering, voxel size=2*2*5 mm, FA=10^o; SENSE=4, TFE factor=14 for two volunteers; SENSE=2, TFE factor=14 for one volunteer; and no SENSE, TFE factor=25 for one volunteer.

After mDixon fat/water separation, water-only images were processed on a pixel-by-pixel basis using custom Matlab routines. Water, fat images and B₀ maps were obtained for each saturation frequency offset. Field inhomogeneity was corrected using an averaged B₀ map from all frequency offsets. CEST maps were generated by integrating MTR_asym in 4 ranges: (i) 0.8-1.2 ppm, (ii) 1.0-1.4 ppm, (iii) 1.2-1.8 ppm and (iv) 3.1-4.1 ppm. The first three ranges are for hydroxyl groups^3,5,6 and the last is for amide groups⁴. Several ROIs were placed manually on normal and/or malignant tissues. The malignant regions were identified by a radiologist and confirmed by US-guided biopsy performed after the MRI scans. The Z-spectra were calculated for each ROI.

Representative images and Z-spectra of the healthy volunteer and a patient volunteer are shown in Fig.1. The malignant tissue displays higher CEST effect than the healthy tissue in all four frequency ranges investigated (Fig.2), with MTR_asym reaching 11% in malignancy in all three hydroxyl ranges and 5% in the amide range. In malignancy, MTR_asym in all the three hydroxyl ranges is higher than the MTR_asym in the amide range, however with larger deviation across patients. By contrast, MTR_asym across all hydroxyl ranges in the healthy tissue display lower effect than in the amide range (with larger deviation too). Hence the MTR_asym difference between the malignant and the healthy tissues is larger in the hydroxyl range (0.8-1.8 ppm) than in the amide range (3.1-4.1 ppm). Interestingly, the Z-spectrum of the malignancy is narrower than in the healthy tissue, as is evident in Fig.1 and further explored in Fig.3.

Overall, the CEST-mDixon sequence is very robust and in most cases, water-fat decomposition leads to homogenous fat removal in the water-only images. We are investigating whether acquiring more echo points will help eliminating the residual artifacts still observed in the areas with large B₀ deviations.

Our human studies demonstrated marked differences in MTR_asym and the width of the z-spectrum in healthy vs. malignant breast tissues. The results suggest that the mDixon-based CEST-MRI has the potential as a robust detection and characterization tool of breast malignancy. Additional work on the delineation of the origins of various CEST signatures, as well as correlation of CEST contrast with different breast tumor types, aggressiveness, and biochemical markers is ongoing.