The International Panel on the Diagnosis of Multiple Sclerosis (MS) incorporated the Barkhof/Tintoré (B/T) magnetic resonance criteria into their diagnostic scheme to provide evidence of dissemination in space of central nervous system lesions, a prerequisite for diagnosing MS in patients who present with clinically isolated syndromes (CIS). Although specific for MS, the B/T criteria were criticised for their low sensitivity and relative complexity in clinical use. We used lesion characteristics at onset from 349 CIS patients in logistic regression and recursive partitioning modelling in a search for simpler and more sensitive criteria, while maintaining current specificity. The resulting models, all based on the presence of periventricular and deep white matter lesions, performed roughly in agreement with the B/T criteria, but were unable to provide higher diagnostic accuracy based on information from a single scan. Apparently, findings from contrast-enhanced and follow-up magnetic resonance scans are needed to improve the diagnostic algorithm.