The effect of several magnetic resonance (MR) variables on the accuracy of volume measurements in phantom objects was investigated by use of an off-line automatic border-outlining and internal area pixel-counting computer program, and an optimal set of imaging variables was identified. Measurements were made of the temporal lobe volumes of a gross fixed brain specimen from MR image data. The range in accuracy was from - 2 to + 7%, and the standard deviation of the difference in right minus left lobe volume measurements obtained from the MR images and those obtained by use of Archimedes' principle was 1 cm3. This volumetric technique was applied to 25 normal persons, most of whom were right-handed. The median ratio of right to left temporal lobe volume was 1.16 (range 0.99-1.23). The nondominant temporal lobe was significantly larger than the dominant. The mean difference (95% confidence interval) between right and left volumes was 7 cm3 (6-9 cm3). This confidence interval was similar to that obtained when the variability within a subject (estimated from the gross fixed brain specimen) was taken into account. Unilateral temporal lobe atrophy, particularly in patients with temporal lobe epilepsy, should be interpreted from MR images with this range of discrepancy in normal left-right size in mind.
(C) Lippincott-Raven Publishers.