Classification of Dynamic Breast MR Image Data - Biomedical Signal Analysis: Contemporary Methods and Applications - page 281

100

75

slice 14 slice 15

50

sa i : 80.38
sv p : 5.81

25

0

slice 16 slice 17

1 2 3 4 5 6

Figure 10.3
Segmentation method I applied to data set #1 (scirrhous carcinoma). The left
image shows the lesion extent over slices #14 to #17. The right image shows the
average time-signal intensity curve of all pixels belonging to this lesion.

and strong contrast-agent uptake is followed by subsequent plateau and
washout phases in the round central region of the lesion, as indicated by
the corresponding CV of cluster #6 in figure 10.5.
Furthermore, clustering results enable a subclassification within this
lesion with regard to regions characterized by different MRI signal time
courses: The central cluster #6 is surrounded by the peripheral circular
clusters #7, 8, and 9, which primarily can be separated from both the
central region and the surrounding tissue by the amplitude of their
contrast-agent uptake ranging between CV #6 and all the other CVs.

Segmentation method III
This segmentation method combines method I with method II. Method
I is chosen for determining the lesions with a super-threshold contrast-
agent uptake, while method II performs a cluster analysis of the identi-
fied lesion.
Figure 10.6 shows the segmentation results for data set #1.

10.2 Results

The computation time for vector quantization depends on the number
of PTCs included in the procedure. The computation time per data set