Objective: The purpose of this study was to evaluate the accuracy of MR imaging for categorizing the configuration of meniscal tears of the knee.
Materials and methods: Fast spin-echo MR images obtained at 1.5 T from 110 patients who had meniscal tears identified at arthroscopy were retrospectively and independently classified by two reviewers into five configurations: horizontal, longitudinal, radial, oblique, and complex. MR imaging categorization was compared with arthroscopic results as the standard of reference. Data were also analyzed with longitudinal and oblique tears combined because these usually are reparable, and with horizontal, radial, and complex tears combined because these usually are not reparable. Interobserver and intraobserver agreements were calculated using kappa coefficients.
Results: At arthroscopy, meniscal tears were categorized as horizontal (n = 44), longitudinal (n = 34), complex (n = 22), radial (n = 11), and oblique (n = 5). Sensitivity, specificity, and accuracy of each reviewer for the reparable tears were 82%, 92%, and 89%; and 59%, 97%, and 84%, respectively. Interobserver agreements were fair between reviewer 1 and the first and second interpretations of reviewer 2 (kappa = 0.25, p < 0.005; and kappa = 0.21, p < 0.05, respectively). Intraobserver agreement was substantial (kappa = 0.71, p < 0.001).
Conclusion: MR imaging was accurate for predicting reparable meniscal tears and was sensitive for the determination of nonreparable tears.