Objective Relapsing remitting Multiple Sclerosis (RRMS) remains an important differential diagnosis of Neuromyelitis optica (NMO). The recently proposed Oxford 2012 diagnostic imaging criteria suggest that analysis of selected data from brain Magnetic Resonance Imaging (MRI) may usefully contribute to the diagnostic process in NMO. When comparing the distribution of brain MRI lesions in patients with NMO vs. RRMS, these criteria demonstrate that patients with RRMS are significantly more likely to have lesions adjacent to the body of the lateral ventricles and in the inferior aspect of the temporal lobes than patients with NMO.

We have tested the Oxford 2012 criteria in analysing MR images of a cohort of well–characterised NMO cases from South Wales and South West England to investigate whether they would improve our ability to differentiate NMO from RRMS.

Methods We obtained 27 brain MRI scans from 24 patients with aquaporin–4 antibody positive NMO. Images were acquired on 1.5 Tesla MRI scanners at 11 radiology departments between February 2006 and December 2012. A single neuroradiologist, blinded to the clinical diagnoses, systematically identified T2 hyperintense lesions on the T2–weighted images, which were then classified according to location and morphology. The Barkhof criteria for dissemination of MS lesions in space, as well as the Oxford 2012 criteria for differentiation of RRMS from NMO were then applied to this data. This allowed us to quantify the sensitivity of both sets of criteria for differentiating NMO from RRMS.

Results T2 hyperintense lesions were identified in 26/27 (96%) brain scans of patients with NMO (median lesion count 9, range 1–60). The majority of lesions were located within the frontal lobes, mostly juxtacortical (median lesion count 3, range 0–28) and in the deep white matter (median lesion count 2, range 0–27). Periventricular lesions were present in 18/27 (67%) scans (median lesion count 1, range 0–9). Temporal lobe, infratentorial, and cerebellar lesions were observed in 4/27 (15%), 6/27 (23%) and 3/27 (11%) respectively.

Application of the Oxford 2012 criteria identified 23/27 (85%) scans which were ‘true negative’ for RRMS compared to 18/27 (67%) using Barkhof's criteria for dissemination of RRMS lesions in space, based on unenhanced brain MRI alone. Of the 9/27 (34%) that fulfilled Barkhof's criteria only 4 were also positive for Oxford 2012 criteria for RRMS.

Conclusion Application of the Oxford 2012 diagnostic criteria in a clinical setting resulted in an 85% specificity for differentiating NMO from RRMS, the most common differential diagnosis compared to only 67% for Barkhof's criteria. This data should be considered in the MRI assessment of patients in whom the diagnosis of NMO is being considered. In particular patients fulfilling Barkhof's criteria should also be assessed with the Oxford 2012 criteria to improve the interpretation of brain MR imaging and its diagnostic accuracy.