Loss of parenchymal volume can be estimated using standard MRIs and a wide range of approaches to image analysis. Different groups have proposed various methods for atrophy assessment, and results from cross-sectional and longitudinal studies have accumulated. Techniques vary according to the extent of automation or user interactivity, and by the structures measured. For example, the degree of operator interaction that is required ranges from completely subjective rating scales to fully automated volumetric calculations. The structures measured range from highly localized measurements of third ventricular width (on the order of 3 mm) to global measurements of whole brain volume (on the order of1,000 cm3). The conceptual approach to measurement also varies widely, from estimating total atrophy since disease onset from a single image, to determining the shift in the edges of a structure between serially acquired images.
One of the most important aspects of CNS atrophy measurements is reliability, because the rate of atrophy in MS patients is relatively slow. Measurement errors can exceed biological change in longitudinal studies, even with follow-up durations of several years. Measurement reliability is usually expressed as reproducibility or inter-rater agreement. This is commonly expressed as a coefficient of variation (CV), defined as the standard deviation of the mean, divided by the mean. Repeated measurements performed on scan-rescan data simulate conditions in serial studies and therefore provide a better estimate of measurement variability than repeated measurements on the same image data.
Was this article helpful?