In recent years, several segmentation techniques have been developed to quantify brain MRI lesion load in multiple sclerosis and are now being used increasingly for both evaluating the natural history of the disease and monitoring the efficacy of treatment. Such techniques have been applied extensively to conventional T2-weighted images; there is less experience to date with gadolinium-enhanced scans and newer MR parameters which are thought to reflect the more destructive aspects of the pathological process. Manual outlining of lesions on T2-weighted images is moderately accurate when performed by an experienced observer and has been validated in large clinical trials, but is very time consuming. Semi-automated intensity-based techniques appear to be more reproducible, but are still slow to perform. Faster techniques with higher reproducibility are still needed. The correlations between conventional T2 lesion load and clinical outcome is strong in patients seen at first presentation with clinically isolated syndromes suggestive of multiple sclerosis, but much weaker in established multiple sclerosis. The application of quantitative assessment of the MR parameters which are more specific for the disabling aspects of the pathological process (i.e. demyelination, axonal loss and spinal cord involvement) may improve the correlation with clinical status.