Accuracy and validity of stereology as a quantitative method for assessment of human temporal lobe volumes acquired by magnetic resonance imaging.

The object of this study was to compare the accuracy and validity of stereology as a method for determining whole temporal lobe volume with the more established technique of semi-automated thresholding and tracing. Ten, fixed, post-mortem human brains, were imaged using a three dimensional (3D) acquisition protocol. The volume of the left temporal lobe, dissected from each brain, was determined by fluid displacement. Each volume was compared to measurements obtained from magnetic resonance images (MRI) of the post-mortem brain using each of the two segmentation methods. Post-acquisition processing was performed using MEASURE software. Three investigators performed each measurement three times using each method, yielding a total of 180 measurements. Stereology took, on average, half the time of thresholding/tracing. Using a clinically acceptable variation for 95% of repeat measures; both intra-observer and inter-observer variation were acceptable for each technique. However, validity, as demonstrated by graphs of agreement against water displacement showed that the "limits of agreement" using stereology were within the acceptable range, while those using the thresholding/tracing technique were not. Quantitative estimates of variation and a graphical representation of the limits of agreement show that stereology is at least as precise as the thresholding/tracing method but is superior in terms of speed and validity. This has broad implications for published estimates of brain region volumes in human diseases such as epilepsy, dementia and other neurodegenerative disorders.