Rationale and Objectives: Quantitative assessment of knee articular cartilage (AC) morphology using magnetic resonance (MR) imaging requires an accurate segmentation and 3D reconstruction. However, automatic AC segmentation and 3D reconstruction from hydrogen-based MR images alone is challenging because of inhomogeneous intensities, shape irregularity, and low contrast existing in the cartilage region. Thus, the objective of this research was to provide an insight into morphologic assessment of AC using multilevel data processing of multinuclear (23Na and 1H) MR knee images. Materials and Methods: A dual-tuned (23Na and 1H) radio-frequency coil with 1.5-TMR scanner is used to scan four human subjects using two separate MR pulse sequences for the respective sodium and proton imaging of the knee. Postprocessing is performed using customized routines written in MATLAB. MR data were fused to improve contrast of the cartilage region that is further used for automatic segmentation. Marching cubes algorithm is applied on the segmented AC slices for 3D volume rendering and volume is then calculated using the divergence theorem. Results: Fusion of multinuclear MR images results in an improved contrast (factor >3) in the cartilage region. Sensitivity (80.21) and specificity (99.64) analysis performed by comparing manually segmented AC shows a good performance of the automated AC segmentation. The average cartilage volume (23.19±1.38 cm3; coefficient of variation COV -0.059) measured from 3D AC models of four data sets shows a marked improvement over average cartilage volume (23.24cm3; COV -0.19) reported earlier. Conclusions: This study confirms the use of multinuclear MR data for cartilage morphology (volume) assessment that can be used in clinical settings. © 2015 AUR.