To compare the diagnostic performance of T2-weighted Dixon, T1-weighted and Short-Tau Inversion Recovery (STIR) MR images for the detection of radiographically occult fractures (ROF) of the hip and pelvis in elderly patients after low-energy trauma.

A total of 22 patients older than 50 years with suspected ROF after low-energy trauma was prospectively included. There were 9 men and 13 women, with a mean age of 80.9 years±12.5 (SD) (range: 52–100 years). T2-weighted Dixon, T1-weighted and STIR MR images were analyzed by 3 independent radiologists blinded to the clinical data and the results of other imaging examinations. Readers separately assessed each series of images for the presence of fractures on a per bone analysis. Diagnostic performance of each reader was compared for Dixon and non-Dixon sequences using contingency tables and McNemar test. Interobserver agreement was evaluated according to the Fleiss-Cuzick's kappa statistics.

The sensitivity of the Dixon sequence in the detection of ROF ranged from 90.9% (20/22; 95% CI: 70.8–98.9%) to 100% (22/22; 95% CI: 84.6–100%). The sensitivities of the non-Dixon sequences in the detection of ROF ranged from 95.5% (21/22; 95% CI: 77.2–99.9%) to 100% (22/22; 95% CI: 84.6–100%). For each reader, there were no statistical differences between combined Dixon and combined non-Dixon images for the detection of ROF (P=0.12, 0.99 and 0.99). Interobserver agreement with T2-weighted Dixon water-only images was significantly lower than that with the STIR sequence (0.70–0.79 vs. 0.87–0.93).

T2-weighted Dixon may be a second-rate alternative to T1-weighted and STIR sequences for the detection of ROF of the hip and pelvis in elderly patients.