Osteoarthritis is a major health issue in western countries. The inherently limited regenerative potential of articular cartilage renders timely diagnosis to be paramount for the possibility of successful conservative treatment. Similarly, a non-invasive method allowing for repetitive and reproducible assessment of cartilage quality would be valuable for interventional studies. It is assumed that even before morphological changes become apparent, early onset osteoarthritis goes along with changes in the biochemical composition of articular cartilage, notably with a decrease in GAG content.¹ Hence, GAG specific MRI techniques, such as sodium imaging and gagCEST bear a high potential of catering to the aforementioned needs.² The aim of this study was to establish an optimized gagCEST method that allows for robust assessment of GAG content in articular cartilage within a clinically feasible measurement time.Seven young and healthy volunteers (mean age 24, range 22-26; m:f = 5:2) were examined on a 7 Tesla MR System (Magnetom, Siemens, Erlangen). The physical integrity of the examined knee joint was assured by orthopaedic examination and assessment of the WOMAC score. Each volunteer was measured twice within one hour using a 28-channel knee array coil (Quality Electrodynamics LLC, Cleveland, OH). A custom-made fixation splint was used to address motion artefacts. A prototype segmented 3D RF-spoiled gradient-echo (GRE) sequence (TE = 3.1 ms, TR = 7.9 ms, resolution = 0.9 x 0.9 x 2.2 mm³, measurement time = 19:39 min) with selective RF presaturation using ten 60 - ms hs2 pulses with variable frequencies (± 10, ± 20 Hz) in the saturation pulse train was applied.³ The images were measured for 19 saturation offsets with a step of 92 Hz around the water resonance. B1 for saturation was set to reach 80% of the specific absorption rate (SAR) limit. Consecutively, offline image registration using an elastic approach was carried out. The calculation of Z-spectra was performed on a pixel-by-pixel basis using spline interpolation of experimental points. Asymmetry of the Z-spectra (MTR_asym) was calculated from integrals over the offset range ± δ = 0.6 – 1.8 ppm relative to the minimum of each individual Z-spectrum. Region-of-interest (ROI) analysis was carried out by one reader. For each region, i.e., weight bearing & non-weight bearing femoral cartilage, trochlear groove, patellar and tibial cartilage, ROIs were placed on three consecutive slices on morphological images and were consecutively transferred to gagCEST maps (Fig. 2). For assessing reproducibility, a two-way mixed intraclass correlation coefficient (ICC) was calculated. Differences in MTR_asym between different regions were assessed using one way ANOVA. Subsequently, least significant distance Bonferroni test was applied.Thin knee cartilage structures and its curved surfaces contribute significantly to B0 inhomogeneity, which hampers uniform saturation of exchangeable OH protons over the entire cartilage volume. Therefore, variation of the nominal frequency of the adiabatic pulses around the offset frequency in the saturation pulse train was performed, which increased the uniformity of saturation in the whole range of resonance frequencies of the OH protons.³ In addition, the custom-made fixation splint successfully addressed motion artefacts and increased patient comfort during measurements. Fig. 1 shows typical MTR_asym values in a healthy volunteer obtained with the optimized protocol. Typical regions of interest selected in various cartilage structured are shown in Fig. 2 and the corresponding mean MTR_asym are given in Fig. 3. There was good agreement between two consecutive measurements as demonstrated by an ICC of 0.82. While MTR_asym values of a similar magnitude were found for weight bearing and non-weight bearing femoral cartilage (p = 1.0), lower values were observed for the trochlear groove (p = 0.006), patellar (p = 0.028) and tibial cartilage (p = 0.006) compared to weight bearing femoral cartilage (Fig. 2).Combined implementation of the modified pulse sequence and optimized fixation allowed to obtain gagCEST maps showing expected distribution of GAG in knee cartilage of young healthy subjects. Significant differences in the mean MTR_asym values were observed between different regions of knee cartilage. The sensitivity to regional differences in the GAG content suggest that the gagCEST method might be useful in examining various pathological conditions in patients.