MR physicists and radiologists interested in the evaluation of biochemical status of articular cartilageA decrease in concentration of glycosaminoglycans (GAG) in articular cartilage is associated with early manifestation of osteoarthritis. However, quantification of GAGs in articular cartilage in vivo is difficult. Several methods have been suggested to be sensitive to the changes in GAG concentration, namely dGEMRIC, T₁ρ mapping, sodium imaging and recently gagCEST. The gagCEST method is promising since it is noninvasive, it does not need extra hardware and the relation of gagCEST values to the GAG concentration have been demonstrated in vitro on naturally degenerated cartilage specimens (1). Similarly, sodium MRI has been validated for the evaluation of GAG concentration in cartilage (2). In our study, we attempted to validate the performance of the optimized gagCEST method for the GAG content evaluation in vivo at 7 Tesla by using corrected sodium signal intensities (cSI) as a reference.Five patients with acute cartilage injury were measured twice with an interval of 8 days between the examinations. The gagCEST and sodium MRI experiments were done on a 7T MR System (Siemens, Erlangen, Germany) using a 28-channel knee array coil and a 15-channel sodium-only array coil (both Quality Electrodynamics LLC, Cleveland, OH), respectively. For gagCEST, a segmented 3D RF-spoiled gradient-echo (GRE) sequence (TE=3.15ms, TR=7.9ms, resolution=0.9×0.9×2.2mm³) was used in combination with presaturation before each segment by means of ten 60-ms adiabatic full passage hs2 RF pulses followed by spoiling gradients. The nominal frequency of the pulses in the train was varied in a range of ± 20 Hz. Nineteen scans with equidistant (93 Hz) offsets in the range of 1680 Hz around the water resonance and a scan without saturation were collected, resulting in a scan time of 20 min. The amplitude of the saturation pulses corresponded to about 80% of the SAR limit. Z-spectra were extracted from registered images and their asymmetry (MTR_asym) was calculated from integrals over the offset range ±∂ = 0.6–1.8 ppm relative to the minimum of each individual Z-spectrum. Sodium images were measured using a 3D vTE-GRE sequence with variable echo times, TR=9.2 ms, TE_min/TE_max =1.22/1.82 ms, FA=51º, bandwidth=100 Hz/pixel, resolution=1.6×1.6×3.0 mm³ and a scan time of 25 min.In the sagittal sodium images and gagCEST maps (Fig. 1), regions of interest were selected in weight-bearing and non-weight-bearing zones of femoral condyles, in patellar and tibial cartilage. For the correlation between cSI and MTR_asym values, mean values of two measurements in the selected regions were used. A good agreement between results of these two methods was observed (Fig. 2). Without considering an outlier, the correlation coefficient was about 0.9. It was also observed that cartilage properties may not be strictly bound to its weight-bearing or non-weight-bearing function and may vary in a relatively broad range.Sodium signal intensities can be affected by hardware imperfections and by partial volume effect due to low spatial resolution of sodium images. In this study, the reported sodium signal intensities were corrected for inhomogeneous B₁ field (3) and for the partial volume effects. On the other hand, the gagCEST values can be affected by the mode of saturation (including direct saturation of water), by the exchange rate and by relaxation times of water and exchangeable protons in the tissue (4). In spite of these limitations, a good agreement between the MTR_asym and cSI values was observed. Because the relation between sodium signal intensities and the GAG concentrations is relatively straightforward and has been proven experimentally, it can be concluded that gagCEST can also be useful in estimating GAG concentration in knee cartilage in vivo. The outlying value in Fig. 2 indicates that corrections for unusual properties of cartilage might be necessary in some cases.