PurposeAxillary lymph nodes (LNs) often present a reservoir for metastatic breast cancer, yet metastatic LN involvement cannot be discerned definitively using diagnostic imaging. This study investigated whether in vivo CEST may discriminate LNs with versus without metastatic involvement.Methods3T MRI was performed in patients with breast cancer before clinically‐indicated mastectomy or lumpectomy with LN removal, after which LN metastasic involvement was determined using histological evaluation. Non‐contrast anatomical imaging, as well as B0 and B1 field maps, were acquired in sequence with three‐point CEST‐Dixon (3D turbo‐gradient‐echo; factor = 25; TR/TE1/ΔTE = 851/1.35/1.1 ms; spatial‐resolution = 2.5 × 2.5 × 6 mm; slices = 10; four sinc‐gauss pulses with duty‐cycle = 0.5, total saturation duration = 701.7 ms; B1 = 1.5 μT; saturation offsets = −5.5 to +5.5 ppm; stepsize = 0.2 ppm; scan duration = 6 min 30 s). The mean z‐spectrum from LNs with (n = 20) versus without (n = 22) metastatic involvement were analyzed and a Wilcoxon rank‐sum test (significance: p < 0.05) was applied to evaluate differences in B0, B1, and magnetization transfer ratio (MTR) in differing spectral regions of known proton exchange (nuclear Overhauser effect [NOE], amide, amine, and hydroxyl) between cohorts.ResultsNo difference in axillary B1 (p = 0.634) or B0 (p = 0.689) was observed between cohorts. Elevated MTR was observed for the NOE (−1.7 ppm; MTR = 0.285 ± 0.075 vs. 0.248 ± 0.039; p = 0.048), amine (+2.5 ppm; MTR = 0.284 ± 0.067 vs. 0.234 ± 0.31; p = 0.005), and hydroxyl (+1 ppm; MTR = 0.394 ± 0.075 vs. 0.329 ± 0.055; p = 0.002) protons in LNs from participants with versus without metastatic involvement.ConclusionsFindings are consistent with a unique metastatic LN microenvironment detectable by CEST‐Dixon and suggest that CEST MRI may have potential for mapping LN metastasis non‐invasively in vivo.