Assess quantitatively whether magnetic resonance imaging (MRI) signal intensity can be used to distinguish cerebrospinal fluid (CSF) leaks in the temporal bone from middle ear effusions.
Retrospective case review.
Tertiary referral center.
Forty-nine patients, 18 with middle ear effusions (MEE), 30 with CSF leaks, and 1 with an MEE on one side and a CSF leak on the other, were evaluated in the study. Primary inclusion criteria for CSF leak patients were operative management of CSF leak with confirmatory diagnosis in follow-up. Primary inclusion criteria for MEE patients were electronic medical record documentation of an effusion with subsequent resolution on follow-up. For all patients, inclusion criteria included MRI imaging with 3D-T2 weighted sequences (3DT2) and fluid-attenuated inversion recovery (FLAIR) sequences performed within 1 year of diagnosis code entry.
Computational analysis of signal intensity of fluid collections in MRI imaging.
Main Outcome Measures:
Sensitivity and specificity of 3DT2 and FLAIR signal intensity in the detection of CSF leak.
For 3DT2 sequences with a chosen normalized signal intensity threshold (CSF leak if ≥ 0.5), sensitivity was 100% (95% CI: 86.3–100) and specificity was 83.3% (95% CI: 51.6–97.9). For FLAIR sequences with a threshold of 1.0 (CSF leak if < 1.0), sensitivity was 77.4% (95% CI: 58.9–90.4) and specificity was 85.7% (95% CI: 63.6–97.0). For a combined test in which a fluid collection was considered CSF if both 3DT2 ≥ 0.5 and FLAIR < 1.0, sensitivity was 76% (95% CI: 54.9–90.6) and specificity was 100% (95% CI: 73.5–100).