Multispectral (MS) analysis was used to determine the ischemic lesion volum
e in the rat brain after permanent middle cerebral arterial occlusion. MS a
nalysis used a four-dimensional MS model consisting of an estimate of the a
verage apparent diffusion coefficient of water (ADC(av)), T2, proton densit
y, and perfusion. Four classification methods were investigated: (a) multiv
ariate gaussian (MVG); (b) k-nearest neighbor (k-NN); (c) k-means (KM); and
(d) fuzzy c-means (FCM). MVG and k-NN classifiers are supervised methods r
equiring labeled training data to characterize the stroke lesion. Unsupervi
sed classifiers (KM, FCM) do not require previous statistics or labeled tra
ining data, resulting in potentially greater clinical usefulness. All MS me
thods provided significant correlation with postmortem findings beyond the
use of ADC(av), alone (partial correlation given the ADC(av) estimate: MVG,
.66; k-NN, .75; HM, .68; FCM, .70). This study demonstrates that MS analys
is provides an improved estimate of ischemic lesion volume over that obtain
ed from ADC alone.