BACKGROUND No study, so far in France, has investigated the diagnosis value of knee MR-arthrography since the recent approval of intra-articular gadolinium use, by this country's healthcare authorities. This study objective is to verify the MR-arthrography superiority on conventional knee MRI, in meniscus and cartilage knee lesions diagnosing accuracy both in regard to sensitivity and specificity. OBJECTIVE MR-arthrography, represents in some pathologic situations, a more accurate source of information than conventional MRI. METHODS Over a 27 months period, 25 patients, scheduled to undergo a knee arthroscopy volunteered, after having been fully informed of the possible interest and risk of the MR-arthrography examination, to participate in this study. Twenty-one of them were finally included since in four cases the surgical indication was not confirmed. The group consisted of 15 males and six females with an average age of 35.7 years. All of them consecutively underwent conventional MRI, MR-arthrography finally followed by arthroscopy. The MRI and MR-arthrograms results were compared to the arthroscopy findings using the nonparametric Kappa test. RESULTS To diagnose meniscal tears, statistical agreement measure for MRI with arthroscopy was good (K=0.69) but not as good as the MR-arthrography/arthroscopy agreement which, by itself was excellent (K=0.84). As a diagnosis tool, the sensitivity and specificity of MR-arthrography (respectively 100 and 89.6%) were much higher than the corresponding values observed in conventional MRI (92.3 and 82.8%, respectively) which nonetheless remain satisfactory. The meniscal tears characterization seemed to be better interpreted using MR-arthrography. As far as the chondral lesions in this series, they were predominantly located on the patellar surface and in the medial femorotibial compartment. For diagnosing the latter, the MRI/arthroscopy agreement was good (K=0.70) but not as good as the MR-arthrography/arthroscopy agreement (K=0.805) which can be rated excellent. The detection sensitivity thus increased by 10% with gadolinium intra-articular injection. However, assessment accuracy of the lesions depth was mediocre, with frequent errors for the intermediary stages. CONCLUSIONS Intra-articular gadolinium injection improved MRI performances for numerous reasons: filling the joint, reinforcing the synovial fluid signal, and enhancing anatomic structures contrast on the T1-weighted sequences images. In this study, MR-arthrography appeared to be superior to conventional MRI in meniscal and cartilaginous lesions diagnosis, confirming the results previously obtained in other countries. In light of these results and other data from the literature, MR-arthrography can be indicated as an alternative to CT-arthrography in various clinical situations: detection of recurrent tears on operated menisci, search for cartilaginous lesions or foreign bodies in the joint space, and preoperative assessment before chondral repair procedures. However, conventional MRI remains the reference examination for studying cartilage, because the low resolution of MR-arthrography limits its performances in quantitative assessment of lesions depth.