inbul 2004 
  
method-A 
1ections to 
r hand, the 
yf method- 
tions were 
1 between 
rror (false 
shown in 
f pixels in 
treated as 
in Figure 
ring their 
ctical use, 
used for 
of their 
is 70. As 
ed as the 
iod under 
mparison 
epresents 
sitive | 
1egative 
  
  
min = 75 
^ 
-min 
than that 
ys larger 
method- 
e 5 data, 
19%); 3, 
iethod-B 
e excess 
hod-C 1s 
error of 
smaller 
that of 
ad us to 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
the conclusion that our proposed method-C has a good effect on 
blood vessel segmentation for MRA data. Several 3-D images 
of the extracted vessel region obtained by method-C are shown 
in Figure 13 and 14. 
  
  
Figure 14. 3-D images of the extracted vessel (data-11) 
4. CONCLUSIONS 
We have developed a branch-based region growing algorithm 
which is designed for blood vessel segmentation for MRA data. 
We examined our segmentation algorithm and its appropriate 
growing conditions for this method. In addition, to perform an 
objective evaluation on the segmentation result, we developed 
the evaluation method based on projection images. We applied 
5 MRA data sets to our segmentation method, and we 
confirmed the validity of our proposed method. 
The biggest problem of our method is its processing time. It 
takes about 5 minutes to get one segmentation result, which is 
ten times longer than conventional region-growing method. 
Most of the time is consumed for branch bifurcation detection. 
It is necessary to optimize the processing for our method to be 
practical. 
The growing condition we have shown is only one instance 
among many. Further discussion on the growing condition is 
still required to utilize this method effectively and to improve 
segmentation reliability. 
Acknowledgements 
This paper is supported in part by Informatics Research Centre 
for Development of Knowledge Society Infrastructure and also 
Intelligent Assistance in Diagnosis of Multi-dimensional 
Medical Images in Grant-in-Aid for Scientific Research on 
Priority Areas. 
  
References 
Masutani, Y., Masamune,K, Dohi,T., 1993. Region - growing 
based feature extraction algorithm for tree-like objects. Proc. of 
VGC'96, Visualization in Biomedical Computing, pp.161-167. 
Masutani, Y., Schiemann,T., Hohne,K.H. 1998. Vascular shape 
segmentation and structure extraction using a shape-based 
region-growing model," Proc. MICCAI'98, pp.1242-1249. 
Frangi, A.P, etal, 2001. Quantitative Analysis of Vascular 
Morphology From 3D MR Angiograms: In Vitro and In Vivo 
Results," Magnetic Resonance of Medicine Vol.45, No.2, 
pp.311-322. 
Fang, L., Wang,Y., Qiu,B., Qian,Y., 2002. Fast Maximum 
Intensity Projection Algorithm Using Shear Warp Factorization 
and Reduced  Resampling, Magnetic | Resonance of 
Medicine .A7(4), pp.696-700. 
Wilson,D.L., Noble,J.A., 1997. Segmentation of Cerebral 
Vessels and Aneurysms from MR Angiography Data, Proc. of 
Information Processing in Medical Imaging, pp.423-428.