Dynamic Cerebral Contrast-enhanced Perfusion MRI - Biomedical Signal Analysis: Contemporary Methods and Applications - page 306

1 2 3 4

5 6 7 8

9 10 11 12

13 14 15 16

Figure 11.1
Cluster assignment maps for the “neural gas” network of a dynamic perfusion MRI
study in a subject with a stroke in the right basal ganglia. Self-controlled
hierarchical neural network clustering of PTCs x ( τ ) was performed by the “neural
gas” network employing 16 CVs (i.e., a maximal number of 16 separate clusters at
the end of the hierarchical VQ procedure). For a better orientation, an anatomic
EPI scan of the analyzed slice is underlaid.

apixel ν characterized by the PTC x ν to

a specific cluster j is based on a minimal distance criterion in the n -
dimensional time series feature space (i.e., ν is assigned to cluster j ), if
the distance

=( x ν ( τ )) ,τ ∈{ 1 ,...,n }

is minimal, where w j denotes the CV belonging
to cluster j . Each CV represents the weighted mean value of all the
PTCs belonging to this cluster.
Self-controlled hierarchical neural network clustering of PTCs x ( τ )

x ν − w j