The volumes of total brain, white matter, gray matter, and cerebrospinal fluid in normal volunteers were determined using a supervised nonparametric feature map segmentation of dual-contrast MR images after correcting for the radio frequency (RF) inhomogeneity. Reproducibility of the segmentation was evaluated in terms of inter-observer and scan-to-scan variations. A comparison with the published results indicates that the scan-to-scan and inter-observer variabilities in the determination of tissue volumes are substantially reduced following the RF inhomogeneity correction.