PURPOSE: To determine the validity and the reliability of T1-weighted three-dimensional gradient-echo magnetic resonance (MR) imaging for quantification of articular cartilage abnormalities of osteoarthritic knees. MATERIALS AND METHODS: Forty-three patients (mean age, 63 years) with knee osteoarthritis (American College of Rheumatology criteria) of the medial tibiofemoral compartment underwent a prospective, cross-sectional study. Knees were examined with a T1-weighted three-dimensional gradient-echo sequence (1.4-mm contiguous sections), with use of a 0.2-T dedicated MR unit, before arthroscopic exploration. The tibiofemoral articular cartilage abnormalities were quantified blindly on both the MR and arthroscopic images with the French Society of Arthroscopy (SFA) score (0-100) and grading scheme (five grades). RESULTS: There was a statistically significant correlation between the SFA-arthroscopic score and the SFA-MR score (r = .83) and between the SFA-arthroscopic grade and the SFA-MR grade (weighted kappa = 0.84). The deepest cartilage lesions graded with arthroscopy and MR imaging showed correlation in the medial femoral condyle (weighted kappa = 0.83) and in the medial tibial plateau (weighted kappa = 0.84). The intraobserver reliability of the SFA-MR score was higher (r = .94) than the interobserver reliability (r = .80). CONCLUSION: Quantification of chondropathy with MR imaging is feasible and well correlated with anatomic cartilage breakdown.