Multivariate analyses of MRI findings for predicting osteomyelitis of the foot in diabetic patients

Abstract

There have been no previous magnetic resonance imaging (MRI) studies using multivariable analysis to diagnose osteomyelitis in patients with diabetic foot. To retrospectively investigate the MRI findings of osteomyelitis in patients with diabetic foot using multivariate analyses. From November 2015 to March 2018, 118 patients who underwent MRI of the foot to evaluate suspected osteomyelitis were included in this study. The patients were categorized into the presence or absence of osteomyelitis. The primary and secondary MRI findings were retrospectively reviewed. To identify independent predictive MRI findings, multivariate analyses with binary logistic regression and receiver operating characteristic curve analyses were performed including all 118 patients and 93 patients presenting decreased T1 signal intensity, respectively. T1 signal intensity, T1 marrow pattern, T1 marrow distribution, T2 signal intensity, concordance of marrow signal intensity, cortical interruption, ulcer depth, abscess, and wet gangrene were significantly different between the two groups (P < 0.05). Multivariate analyses indicated that fluid equivalent T2 signal intensity, deep ulcer, and confluent T1 marrow pattern were major factors associated with osteomyelitis. The area under the curve of predicted probabilities for the combination of these factors was 0.799 across all 118 patients and 0.761 across 93 patients with decreased T1 signal intensity. Confluent T1 marrow pattern is a reliable finding to suggest osteomyelitis in patients with diabetic foot. In addition, fluid equivalent T2 signal intensity and deep ulcer are important findings that may suggest osteomyelitis, irrespective of T1 signal intensity change.