Observer experience affects MRI accuracy for dog knee meniscal tears

Abstract

Low-field MRI (lfMRI) has become increasingly accepted as a method for diagnosing canine meniscal tears in clinical practice. However, observer effects on diagnostic accuracy have not been previously reported. In this study, 50 consecutive stifle joints with clinical and radiologic evidence of cranial cruciate ligament insufficiency were investigated by lfMRI and arthroscopy. Fifteen observers who had varying levels of experience and who were unaware of arthroscopic findings independently reviewed lfMRI studies and recorded whether lateral and medial meniscal tears were present. Diagnostic accuracy (sensitivity, specificity, positive (PPV) and negative predictive value (NPV)) was determined for each observer and median values were calculated for all observers, using arthroscopy as the reference standard. Interrater agreement was determined based on intraclass correlation coefficient (ICC) analysis. Observer level of experience was compared with diagnostic sensitivity and specificity using correlation analysis. Based on pooled data for all observers, median sensitivity, specificity, PPV, and NPV for lfMRI diagnosis of lateral meniscal tears were 0.00, 0.94, 0.05, and 0.94, respectively. Median sensitivity, specificity, PPV, and NPV for medial meniscal tears were 0.74, 0.89, 0.83, and 0.79, respectively. Interrater agreement for all menisci was fair (0.51). Menisci were less consistently scored as having no tears (ICC = 0.13) than those scored as having tears (ICC = 0.50). No significant correlations between observer experience and diagnostic sensitivity/specificity were identified. Findings indicated that the accuracy of lfMRI for diagnosing canine meniscal tears was poor to fair and observer-dependent. Future studies are needed to develop standardized and widely accepted lfMRI criteria for diagnosing meniscal tears.