Differentiation of inflammatory from degenerative changes in the sacroiliac joints by machine learning supported texture analysis

PurposeTo compare the diagnostic performance of texture analysis (TA) against visual qualitative assessment in differentiation spondyloarthritis (SpA) from degenerative changes sacroiliac joints (SIJ).MethodNinety patients referred for suspected inflammatory lower back pain rheumatology department were retrospectively included at our university hospital institution. MRI 3 T lumbar spine and SIJ was performed with oblique coronal T1-weighted (w), fluid-sensitive fat-saturated (fs) TIRM fsT1w intravenously contrast-enhanced (CE) images. Subjects divided into three age- gender-matched groups (30 each) based on definite clinical diagnosis serving as reference standard either degenerative, or no SIJ. rated qualitatively by two independent radiologists quantitatively region-of-interest-based TA 304 features subjected to machine learning logistic regression randomized ten-fold selection training validation data. Qualitative quantitative results evaluated compared standard.ResultsAgreement radiologist’s diagnose fair both readers (κ = 0.32 0.44). ROC statistics revealed significant outperformance ratings SpA remainder (AUC 0.89 vs. 0.75), 0.91 0.67) TIRM-positive (i.e. bone marrow edema) cases 0.95 0.76). T1w-CE images most important discriminator detection SpA.ConclusionsTA is superior Intravenous CE-images increase yield TA.