MRI‑based scoring system to predict spondylodiscitis: The SPONDY-Score

Objective To evaluate the diagnostic performance of a MRI-based scoring system to differentiate endplate edema related to spondylodiscitis and degenerative spine disease (DSD). Methods This retrospective study included 126 patients with spondylodiscitis and 162 with DSD. MRI scans were reviewed by two raters to report 12 features. Excluding collections, which were deemed pathognomonic of spondylodiscitis, the top five most significant features were selected and used for logistic regression. The dataset was randomly split into training and testing sets (75:25). We used the coefficients deriving from logistic regression to construct a clinical score (SPONDY-Score), which would assign said points to each included variable and chose the best threshold for such score. Reproducibility of MRI features was evaluated by calculating Cohen κ and raw concordance. Results The SPONDY-Score involved the assignment of 1 point for the involvement of posterior elements by bone edema, 15 points for vertebral bodies destruction, 5 points for the presence of paravertebral edema, −4 points for the presence of T2 hypointensity of the disc, 1 point for the presence of non-anatomical T2 hyperintensity of the disc. The best threshold was set at 5 points. When the SPONDY-Score was applied to the testing set, combined with the presence of paravertebral collections as indicator of spondylodiscitis, the model obtained 89 % accuracy, 95 % sensitivity and 85 % specificity. Regarding inter-observer reproducibility, κ values and concordance ranged from 0.61 to 0.90 and from 85 % to 98 %, respectively. Conclusion The SPONDY-Score demonstrated excellent diagnostic performance, providing an accurate and standardized method for clinical decision-making.