Evaluating and validating two methods for estimating brain structure volumes: tessellation and simple pixel counting.

Developments in imaging technology have made three-dimensional visualization of internal brain structures possible with excellent resolution. Since improved visualization implies improved measurement, these advances hold promise to more accurately measure the volumes of internal structures. As new technologies and techniques emerge, evaluating the relative benefits of measurement methods becomes necessary. We compared and evaluated two methods of estimating volumes from images of brain structures. One method counted pixels within a region of interest, while the other method tessellated the surface between tracings on adjacent slices. Our study assessed both measurement error for true phantom volumes and method disparity for in vivo structures in a randomly selected sample of subjects (n = 100). For our comparisons, we focused on the temporal lobe, ventricular system, and hippocampus. Bias, independence of measurement errors and maximal discrimination of individual differences are properties that are relevant to validating and evaluating measurements of cerebral structure. Pixel counting proved to be the more robust of the two methods, being less sensitive to nuisance-interactions between size of object, shape, and slice thickness. Clinical and research applications of imaging techniques may have distinctive but overlapping needs when evaluating and validating new developments in imaging.