Diagnostic performance for erosion detection in sacroiliac joints on MR T1-weighted images: Comparison between different slice thicknesses

Purpose: To assess the effect of slice thickness on the diagnostic accuracy of erosion detection at MR T1-weighted images (T1WI) of the sacroiliac joints (SIJ) in adult patients suspected of sacroiliitis. Method: Patients aged 18-60 years with clinical suspicion of sacroiliitis were enrolled. All patients underwent CT and 3 T MRI of the SIJs on the same day. CT at 1 mm slice thickness, semi-coronal spin echo T1WI sequences with four different slice thicknesses (2, 3, 4 and 5 mm) were obtained. For scoring erosions, each SIJ was divided into four quadrants. Presence or absence of erosions was scored on T1WI sequences by two independent readers blinded to other data. Inter-reader agreement was assessed using kappa statistics. Diagnostic accuracy of MRI for erosions at each slice thickness was evaluated vs. consensus CT as reference standard, using area under the receiver operating characteristic curve (AUC). Results: Fifty-three patients (23 men, 30 women, mean age, 39.0 years +/- 10.2) were included. Inter-reader agreement for erosion score on all T1WI sequences was moderate (kappa value 0.54 to 0.60). With increasing slice thickness, both the recorded total number of erosions and sensitivity for erosion vs. CT decreased. The AUC were significantly higher for 2 mm and 3 mm T1WI than for 4 mm and 5 mm T1WI. Conclusions: The diagnostic accuracy of T1WI for erosion detection vs. a CT reference standard is affected by slice thickness. Thinner slices (2 or 3 mm) had significantly higher diagnostic accuracy than thicker slices (4 or 5 mm).