Magnetic resonance imaging (MRI) is an important paraclinical tool for the diagnosis of multiple sclerosis (MS) and for monitoring its disease course. The efficacy of most of the available MS disease-modifying treatments has been tested in clinical trials where MRI-derived quantities served as primary or secondary outcome measures. However, conventional MRI measures (i.e., the number and volume of contrast-enhancing, the volumes of T2-hyperintense and T1-hypointense lesions and the assessment of brain volume changes) are limited in terms of pathological specificity and, as a consequence, are modestly correlated with clinical measures of disease activity and have a modest prognostic value as predictors of MS evolution. In the present review, we discuss the main factors potentially responsible for the so-called 'clinical MRI paradox' and how modern quantitative MR-based techniques might contribute to, at least partially, overcome it. The lessons learned from MS trials suggest that future applications of MRI to assess MS evolution should rely upon the use of composite measures thought to reflect the various components of the disease, as well as on study protocols specifically designed on the individual trial characteristics.