Interobserver variability of dynamic MR imaging of the temporomandibular joint

This study investigated the interobserver variability of dynamic magnetic resonance (MR) imaging of the temporomandibular joint (TMJ). Sixty-two patients with internal derangement of the TMJ (124 TMJs) were investigated with a 1.5-T MR imaging system during physiological opening and closing of the mouth. Two readers evaluated independently the quality of the dynamic examination (Q: nondiagnostic, diagnostic, optimal), condylar motion (CM: limited, suboptimal, optimal), condylar orientation (CO: in-plane, through-plane shift), disc visibility and movement (DV: visible, nonvisible; DM: normal, reducing, nonreducing dislocation) and joint effusion (JE: present, absent). For each TMJ, the condylar path was measured by tracing the position of the condyle in the frames of the dynamic acquisition. Agreement between the two readers was assessed with Cohen's Kappa and the Bland-Altman method. Interobserver agreement was almost perfect for Q (nondiagnostic, diagnostic, optimal: 0.8%, 4.9%, 94.3%; kappa=1), CM (limited, suboptimal, optimal: 14%, 26.4%, 59.5%; kappa=0.84) and DV (visible, nonvisible: 100%, 0%). Substantial agreement was found for DM (normal, reducing, nonreducing: 66.1%, 14.8%, 19.1%; kappa=0.64) and JE (present, absent: 41.3%, 58.7%; kappa=0.67). Moderate agreement was found for CO (in-plane, through-plane shift: 94.2%, 5.8%; kappa=0.41). As for the condylar path, the means of the percentage differences and limits of agreement (LA) were -3% (LA: -34.5%, 28.3%) on the right and -1.2% (LA: -35%, 32.6%) on the left. In dynamic imaging of the TMJ, qualitative assessment of condyle-disc movement and joint effusion is minimally dependent on the reader's evaluation. Measurement of the condylar pathway shows an interobserver variability of +/- 30%.