Objective: To develop an MRI based diagnostic algorithm to distinguish AQP4 positive neuromyelitis optica spectrum disorder (NMOSD) from multiple sclerosis (MS).

Background: NMOSD (MS) are inflammatory diseases of the CNS that share clinical and MRI features making distinguishing these conditions difficult. We have conducted a cohort comparison analysis of imaging data. The results were used to develop predictive models for distinguishing NMOSD and MS.

Design/Methods: A literature search to identify MRI lesions and features associated with NMOSD and define those lesions was undertaken. MS lesions were identified from published diagnostic criteria. Cases of AQP4 seropositive NMOSD were collected from across Australia and New Zealand using the 2015 International panel for NMO diagnosis criteria. Age and sex-matched MS cases were collected through the same centres. MRI were reported by at least two blinded reviewers. Relative frequencies of lesions were compared using odds ratios and predictive models were developed using scores and machine learning.

Results: MRI brain or spine was available for 166/168 (99%) of NMOSD and MS cases. Lesions showing greatest association with NMOSD were lesions of the spinal cord that were longitudinally extensive (OR=203), “bright spotty” (OR 93.8), whole (axial; OR=57.8) or gadolinium (Gd) enhancing (OR=28.6), optic nerve lesions that were bilateral (OR=31.3) or Gd-enhancing (OR=15.4), and nucleus tractus solitarius (OR=19.2), periaqueductal (OR=16.8) or hypothalamic (OR7.2) brain lesions. Lesions with greatest association with MS were ovoid (OR=0.029), Dawson’s fingers (OR 0.031), basal corpus callosum (OR=0.058), periventricular (OR=0.136), temporal lobe (OR=0.137) and T1 black holes (OR0.154). A score-based algorithm and a decision tree determined by machine learning accurately predicted more than 85% of both diagnoses using first available imaging alone.

Conclusions: We have confirmed NMOSD and MS specific MRI features and combined these in predictive models that can accurately identify more than 85% of cases as either AQP4 positive NMOSD or MS.