To compare diagnostic performances for cholesteatoma diagnosis of incremental MRI protocols including non-echo planar diffusion-weighted imaging (DWI) performed on 3T and 1.5T scanners.

Thirty-nine patients with suspected cholesteatoma underwent 3T and 1.5T non-echo planar DWI and additional unenhanced T1-, delayed gadolinium-enhanced T1- and high-resolution T2-weighted standard acquisitions. Patients either underwent surgical tympanoplasty (n=21) or close clinicoradiological follow-up (n=18). Four radiologists independently and prospectively interpreted two incremental MRI protocols, differing in the magnetic field strength of the diffusion-weighted acquisition and comprising the three standard sequences. At each step, diagnostic performances were expressed as sensitivity, specificity, positive predictive value, negative predictive value and accuracy.

Forty middle ear lesions including 21 cholesteatomas were identified. Univariate and multivariate analysis did not demonstrate significant reader, sequence addition or DWI magnetic field effect on diagnostic performances. Concerning non-echo planar DWI alone, sensitivity, specificity, positive predictive value, negative predictive value and accuracy ranged between 90.5–100%, 68.4–100%, 76.9–100%, 90.0–100% and 82.5–95.0, respectively.

Non-echo planar DWI for cholesteatoma diagnosis can be performed on 1.5T or 3T scanners indifferently. High sensitivity and negative predictive value and relatively lower specificity and positive predictive value are achieved by a single non-echo planar DWI protocol.