The Application of Cellular Automaton in Medical Semiautomatic Segmentation - Cellular Automata in Image Processing and Geometry - page 167

Fig. 9.5 Segmentation difference between MSRM and RCA method (highlighted by red
regions)

Ta b l e 9 . 1 Comparison of RCA and MSRM

Image Method Time (s) TPR(%) FPR(%)

Woman RCA 22.77 97.75 1.26

MRSM 30.68 94.68 4.49

Church RCA 25.63 98.07 0.92

MRSM 37.28 96.32 1.34

Fish RCA 19.12 96.11 2.87

MRSM 30.42 95.79 2.93

9.3.3 Volume Cellular Automaton Segmentation

According to traditional 2D interactive method, the straightforward 3D expansion
needs laying seeds in each slice, obviously the manual work is hard and not rea-
sonable. In medical system, user interaction is often performed in the selected MPR
(Multi-Planar Reformation) image planes to indicate the position and main features
of the desired 3D objects. The contours should be extracted by efficient methods
from volume data. This operator may satisfy most expert clinician's habits. Interac-
tions in different MPR image planes are demonstrated in Figure 9.6.

9.3.3.1 Brute-Force Algorithm

For the convenience of description, we select a series of MRI (512 56) to

demonstrate this strategy. Samples of MRI series are shown in Figure 9.7. The
goal is to track the 3D brain soft tissue with complicated shape and process the

× 512 ×