Magnetic resonance imaging of amputated human knees was performed to determine optimal sequences for depicting articular cartilage. 24 knees were examined with eight different sequences in a 1.0 T imager. Each cartilage lesion was graded from 1 to 4 (Outerbridge staging system). The results of each sequence were compared with the macroscopic findings and statistically tested against each other. The FLASH sequence (TR = 50 ms) with combination of flip angle of 40 degrees and echo time of 10 ms and the FISP sequence (TR = 40 ms) with combination of flip angle of 40 degrees and echo time of 11 ms were best for depicting cartilage structure and internal detail. There was no significant difference between fat-saturated three-dimensional FLASH (FS-3D-FLASH) and FS-3D-FISP (p = 0.05). These FS-3D sequences were significantly better than sequences without fat saturation (p = 0.05). There was no significant difference between magnetization transfer (MT) 3D-FLASH, MT-3D-FISP and 3D-FISP. All 3D sequences showed significantly (p = 0.05) better results than spin echo or fast spin echo sequences. The T1 weighted SE pulse sequence was significantly (p = 0.005) better than the T2 weighted TSE sequence. Fast T2 weighted spin echo was not suitable for early and accurate detection of cartilage lesions.