How accurate are measurements on MRI? A study on multiple sclerosis using reliable 3D stereological methods.

Unbiased stereological principles for quantifying plaque volume and number in multiple sclerosis (MS) are described, and practical problems with their implementation are discussed. Plaque volume was estimated using stereological methods on 3 mm brain magnetic resonance imaging (MRI). Volume estimates made from thick sections are biased. For convex particles as plaques, bias can be corrected by excluding the slice of maximal area from the estimate. Comparing the corrected with the uncorrected results, bias varied between 15 and 200%, depending on object size. By analogy, for the estimates of plaque, bias varied between 15 and 90%. The use of 1 mm slices reduces bias to a value close to zero and should be preferred when precise plaque measurements are required. An unbiased three-dimensional counting rule-the disector-was used to estimate plaque number. The coefficient of error of the estimates was calculated. Stereological methods applied to MRI provide efficient and reliable estimates of MS plaque load.