OBJECTIVES To evaluate quantification of early fatty infiltration in supraspinatus muscles with magnetic resonance (MR) imaging using a T2*-corrected multi-echo 3D-gradient-echo Dixon-based sequence (multi-echo Dixon) and compare it to proton-MR-spectroscopy. METHODS Sixty subjects (mean age 46 years, 41 men) with good supraspinatus muscle quality on 1.5 T MR imaging were included. Fat percentage (FP) in the supraspinatus muscle was quantified using a multi-echo Dixon compared to single-voxel MR spectroscopy as reference standard. In 18 subjects the multi-echo Dixon was repeated to assess test-retest reliability. Measurements based on multi-echo Dixon were performed by two independent readers by placing regions-of-interest (ROIs) in the supraspinatus muscle corresponding to the MR-spectroscopy voxel. Intraclass and concordance correlation coefficients (ICC/CCC) were used for statistical analysis. RESULTS Test-retest reliability was substantial for reader 1 (ICC = 0.757) and almost perfect for reader 2 (ICC = 0.873). Inter-reader reliability for multi-echo Dixon was almost perfect (ICC = 0.893, P < .0005). Mean FP in all 60 subjects with multi-echo Dixon was 3.5 ± 1.6 for reader 1, 3.7 ± 1.8 for reader 2, and 2.8 ± 1.4 with MR spectroscopy. Correlation between multi-echo Dixon and MR spectroscopy was moderate (CCC = 0.641). CONCLUSION The multi-echo Dixon sequence is a reliable method and comparable to MR-spectroscopy for quantification of low levels of fatty infiltration in the supraspinatus muscle. KEY POINTS • Multi-echo Dixon for low fat quantification in muscles is reliable. • Multi-echo Dixon low fat quantification is comparable to single-voxel MR spectroscopy • Multi-echo Dixon detects substantial differences in fatty infiltration within Goutallier 0-1.