Knowledge-Based
Fuzzy Shape Detective Algorithm
for Segmenting Human Brain MR Images
Syoji Kobashi, Katsuya Kondo, Yutaka Hata, Mieko Matsui(*
*)Brain Function Research Institute, Inc.,
Knowledge-Based Intelligent Information Engineering Systems & Allied Technologies, pp. 600-603, Sep. 2002
This paper gives a new shape-based segmentation algorithm. This algorithm does not find the contour like the conventional active contour models but detects the representative line. The detection is done by inferring the most suitable direction to extent the representative line according to the expert's knowledge. The inference is performed with fuzzy if-then rules derived from expert's knowledge. We apply this algorithm to segmenting the lateral ventricles from human brain 3-D MR images. The segmentation accuracy is evaluated by comparing the automated volumetry with manual volumetry by a physician.