data set T = {(xv,yv)\ve {1,...,p}} required for supervised learning is provided by interactive labeling of a small representative subset of the data by a human observer. This is performed by manual contour tracing of regions that can clearly be attributed to one of the tissue classes without ambiguity. Figure 8 shows an example for regions manually labeled as "gray matter," "white matter," and "CSF." If a training vector x^ can be assigned to the tissue class X, the corresponding component of the target vector yv is set to one, whereas all the other components are set to zero.

0 0

Post a comment