Tj. Grabowski et al., Validation of partial tissue segmentation of single-channel magnetic resonance images of the brain, NEUROIMAGE, 12(6), 2000, pp. 640-656
We describe and evaluate a practical, automated algorithm based on local st
atistical mixture modeling for segmenting single-channel, T1-weighted volum
etric magnetic resonance images of the brain into gray matter, white matter
, and cerebrospinal fluid. We employed a stereological sampling method to a
ssess, prospectively, the performance of the method with respect to human e
xperts on 10 normal T1-weighted brain scans acquired with a three-dimension
al gradient echo pulse sequence. The overall kappa statistic for the concor
dance of the algorithm with the human experts was 0.806, while that among r
aters, excluding the algorithm, was 0.802. The algorithm had better agreeme
nt with the modal expert decision (kappa = 0.878). The algorithm could not
be distinguished from the experts by this measure. We also validated the al
gorithm oil a simulated MR scan of a digital brain phantom with known tissu
e composition. Global gray matter and white matter errors were 1% and <1%,
respectively, and correlation coefficients with the underlying tissue model
were 0.95 for gray matter, 0.98 for white matter, and 0.95 for cerebrospin
al fluid. In both approaches to validation, we evaluated both local and glo
bal performance of the algorithm. Human experts generated slightly higher g
lobal gray matter proportion estimates on the test brain scans relative to
the algorithm (3.7%) and on the simulated MR scan relative to the true tiss
ue model (4.4%). The algorithm underestimated gray in some subcortical nucl
ei which contain admired gray and white matter. We demonstrate the reliabil
ity of the method on individual 1 NEX data sets of the test subjects, and i
ts insensitivity to the precise values of initial model parameters. The out
put of this algorithm is suitable for quantifying cerebral cortical tissue,
using a commonly performed commercial pulse sequence. (C) 2000 Academic Pr
ess.