MR imaging of fatigue stress injuries to bones: intra- and interobserver agreement

The aim of this study was to determine the validity of MR imaging (MRI) in the assessment of stress-related injuries to bone. MR images of 50 military recruits (8 females and 42 males; 18–27 (mean 20) years of age) were retrospectively evaluated twice for stress injuries to bone by 4 radiologists (2 musculoskeletal radiologists, 2 radiology residents). Coronal T1-weighed (T1W) and STIR images, as well as axial and coronal T2-weighted (T2W) fat-suppressed images were taken using a 1.0T scanner. Rates for sensitivity, specificity, and accuracy of MRI of the stress-related injuries were calculated. Intraobserver and interobserver agreement was determined with kappa statistics. Rates for MRI sensitivity were 27–96%, for specificity 65–100%, and for diagnostic accuracy 58–97%. Lowest rates were seen when reading T1W images and highest when reading STIR images. Readers showed moderate to excellent intraobserver agreement (κ 0.75–0.95). Interobserver agreement was fair to excellent (κ 0.41–0.91), and the lowest values were seen in the interpretation of T1W images. Normal findings could be differentiated from various grades of stress injury to bone. MRI is a valid means of revealing the presence of stress injuries to bone and their staging. Observer agreement is good to excellent when using T2W images and STIR images, while T1W images are of lesser value.