ѧ
JOURNAL OF SOFTWARE
1999 19 3  Vol.19 No.3 1999



ĽʽͼָӦ

ժҪʽͼָܹӵͼõ˹㷺о.һֻڴģ͵Ľʽͼָ.ضܵģ,ͨƽͻ㱨ʵߵĽ.ôָΪɲ֮һ.÷ҽѧͼָеӦȡ˽ϺõĽ.
ؼʡͼָ,,,ҽѧͼ.
ͼšTP391
Region Feature-based Interactive Segmentation Method
and Its Application in Medical Image Analyzing
LIU Ning-ning TIAN Jie
(Institute of Automation The Chinese Academy of Sciences Beijing 100080)
Abstract The interactive segmentation techniques attract more and more attention in that these techniques can handle complicated images. In this paper, an interactive image segmentation method is proposed based on proxy model. The proxy is a task-specific module which carries out certain function. Communications between an operator and computer is realized by means of the control interface and the report interface of proxy. The proposed proxy is constructed on the basis of region features of a to-be-segmented region in an image. This method is applied to solve the segmentation problem in medical images and the result shows that this method is effective in such complicated image as medical image.
Key wordsImage segmentation, proxy model, region feature, medical image.
ͼָǼͼеһ,䱾δõԲ,ûһͨöЧķָܹ㲻ͬĿĵҪ.һҽѧͼԵøΪͻ.ҽѧͼǷӳ֯ĸͼ,Ϣ,ȱٿɹʹõܹ̻֯ѧģͣ1.ͼзָ,õĽ淽Ч.Ŀǰ,ͼָ˶ַ̽,ף24ݵȵ.Щһ̶Ͻ˽ʽָ˼.ҽѧͼصʵӦõҪ5,һںϷָϢĽʽģ,ڷָߵ.Ǹݸģ,MRIͼΪʵ˳ʵ֤,ʵ,һֵֵ̽ķ.
1 
ָ,ǿԽͼָΪӹ:Ŀʶλ̺Ŀȡ,ȱһ.淽ǽָȫģͻ,ɼ,Ҫ˵Ĳ.,ʵָĽ׼ȷԽϲ.ԶָһԲͬоͬķ,кܴľ.ҽѧͼ,ڲ֯䶯Լص,ʹԶָ˸ӵ޲,ӶԶָʧ.,Զָǻģ͵ķָ,ҽѧͼ,޴ӻÿڷָĿĵģ,ӶעԶָڶάݵҽѧͼеʧ.
Ŀ궨λȡĽǶȳ,ǿԷһص:óĿʶλ,㷨ھȷؽĿȡ.˿׵طָȤĿĿռλ,ҶԵظĿĴСͼ״.ֶǼ߱,תΪӦõѧģǲ,ͼָԲѵ.Ȼ,ѧģ;ȷȡĿĪ.,˵л뵽ͼָ,ܵõĽ,ҪʱĿĶλ,˸ʵʾ,ȷļ㴦ɼִ.Դﵽȡ̵Ч,пܵõķָ.ͼָʹЧںϽȥһ˼Ĺؼ.Ϊ˲ԶָĹҪһĶ,:Լĳ̶.ǳож׼:һڷָΪûṩЧĿƷ;һǾ˵Ĳʱ,ȷʵʱʵ.
1.1 ڴĽģ
ڱУʵ˴Ϊ˻ľģ,ͨ˵лؽ.Ƕһɼִеġضܵļģ,ɾȷļ.ܹɼ,ұõ˻ͨѶĽӿ.ͨƽͻ㱨ʵߵͨѶ.ṹͼ1ʾ.ĳһɹ,ʱ߻㱨䵱ǰĹ״̬,һ׶εʱ,ɽʱύ.߿Ըݴ״̬׶ʱƲ,ָµĹ.ڴִоļʱ,߲и,ֻǸݴ״̬ɽȷĹ״Ƿ,ԼύǷԽ,ӶȷһĹĿ.
ͼ1ʾ,ͨƽΪṩָʾ,ͨ㱨߻㱨д״̬.ڽ׶ɺ,׶ɽͨлύ߽.ڱУ˾̬,߽ݴĽ׶ɽпƲԵ޶,ָһ.ڽṹͼ,˵,ҲԿ,ģ;з.,ǿԽͼָеĿ궨λһύ,ȷļɴ,ҿʱƴĲ.

ͼ1
1.2 Ĺ
ģеĹؼǴĽ,ĴھоҪ.ڱ,ʵںϷָľ̬ģ.ģ:ģѰģ.ģضԼȡ;ѰģɱԵŻ.ͨʵֽͨѶ.˵ԭ,߿׵жָĿռλ,Ӷʼͼ2ڷʾķΧ.ڷΧеָָĳһصΪĹ,ԸõΪʼ,ģȽ,ﵽһҪʱ,ȡ,ȻȴָѰģ.ѰģĹʽͼ3ʾ,


ͼ2ͼ3 
ͨԤȴԵָ·,ʹݸóʼ·б߽,õڸ·µĽ·,ȷýǷԽ.ѰģĹͼ4ʾ,ģΪԤָı߽㼰ڵص,ֲָĽű߽,ҽ,Խ,ʼ·µļ,ֱõķָΪֹ. 

ͼ4
ڱ,Ƕ߽ԪΪ֮,ĳһ,ܵı߽Ԫ8,ֱΪk.45,k=0,1,2,3,4,5,6,7.,̻߽ԪǷܹΪĿԵ,ʵв÷ָͳƽҶֵƽҶȾֲһݶȱ仯Ϊ̻߽Ԫص.:
(1) ƽҶPiΪҶֵiĸ,ضĻҶƽֵΪ,,һƽҶֵΪ,MΪضҶֵ.
(2) ƽҶȱ仯 ضĻҶƽ仯Ϊ,һΪ.
,ǵԵ,ݶȵı仯ȱȽ,ʵв˾ֲݶȱ仯Ϊ,ǶΪgrad=((max(gi)-gi))/max(gi),giΪָ·ϵһϵصݶֵ,iʦ,Ϊָص㼯.ֵתΪָ꺯жϾֲرǷָԵĿ,ɴָ꺯ѰҷָԵ,ΪĹ.ָ꺯Ϊ,ʽfΪ,cΪת,wΪȨֵ.ڱУѡfΪƽҶȾֵƽҶȱ仯ֲݶȱ仯.ڻҶƽֵҶȷ,ѡתΪ˹,Էõ:ԽС,G(x)ֵԽ;ھֲݶ,ѡԺΪת,˵õR=1.G(,)+2L(grad),ע⵽ݶʾ,ݶֵԽ,gradֵԽС,ڴ,轫Ŵ,20.ԺѡΪL(x)=.grad,iΪָص㼯еԪ.
1.3 ʵ
̷ָΪ׶.1׶Ϊָ׶.ڸý׶,ɿԤָѡһ㣨ͼ2ʾ,ȻݸϢ.Ӧ.ԲָԭʼΪĵcore,coreΪĵ,õ55,ƽҶȦ=,ЦΪcoreΪĵ55,iΪصĻҶֵ.ڱ,ֵth=-core,ֵ׼жԳΪָڵص.㷨:踳ͼ{g(i,j)}ÿnάF(i,j)=f1(i,j),f2(i,j),...,fn(i,j)T,ĵΪC(a,b)=core1(a,b),core2(a,b),...,coren(a,b)T, ΪM(i,j;a,b)=C(a,b)-F(i,j)=nkcorek(a,b)-fk(i,j),ڱУΪصĻҶֵ.,ԸݿߵУµ¦=,ǿɱ.ֹͣʱ,ȴߵ,ԾǷֹͣ.㷨α:,ĵ趨Ϊָص,ԺĵΪ,55,㦨ĻҶȾֵ,ֵΪ.
BEGIN
Step 1:
core=Selected-key-point();
=Build-neighbors-of (core,5);
=Mean-of-();
th=-core;
End of Step 1
Ƕ3:(1) ָ򼯺R_Set,ڱеõķָص;(2) 㼯Cal_set,еԪرǷڷָ,ǵķ,Cal_setеԪǵ4,4;(3) ʱtmp_set,ڸüбµĺѡĵ,Cal_setеԪظݸüϵص.Cal_setеԪԤָ,򽫸õ뵽R_Set,ͬʱɸõ4ص㲢Ҽ뵽tmp_set,ȴһĴ;Cal_setеĵ㲻Ԥָ,򽫸õtmp_setɾ,뵽final_set.мfinal_setʱҰ㡱(νҰ㡱ϵָвҪص),ⲿص.㷨㵽tmp_setԪʱ,ֹͣ.㷨Ƚгʼ:
BEGIN
Step 2:
Initialize_set (R_Set);
Initialize_set (tmp_set);
Initialize_set (Cal_set);
End Step 2
END BEGIN
ʼ,R_Setҽһصcore;㼯Cal_setʼcoreΪĵõcore4ص,Cal_set={core(i,j-1),core(i,j+1),core(i-1,j),core(i+1,j)}.tmp_setΪ.ֵ׼жϼ㼯еصǷָ,,򽫸ص뵽R_Set,ͬʱɸõ4ص,ҽ4صtmp_set,ȴһ.õ㲻ڷָ,򽫸õӼ㼯ɾ,뵽final_set,ȴ;μ㣬ֱCal_setΪΪֹ.Ȼʱtmp_setеĸصΪµĵ,ص4,Cal_set,ظϲ,ֱtmp_setٴκصΪֹ.ʱ,뱣֤tmp_setظԪ.ٴfinal_set еҰ,Ա֤еҰܹõȷ.㷨:
BEGIN
While (Cal_set not empty)
{poi=select_poi_from (Cal_set);
ifpoi-̡th then
add_poi_to (R_Set);
Net4=Build_neighbor_of (poi);
add_Net_to (tmp_set);
Delete_poi_from (Cal_set);
else
Delete_poi_from (Cal_set);
add_poi_to (final_set);
}
While (tmp_set NOT EMPTY)
{poi=Select_poi_from (tmp_set);
Nei4=Build_neighbor_of (poi);
if pixel(NeiR_Set) then Delete_poi_from (Nei);
add_Nei_to (Cal_set);
While (Cal_set not empty)
{poi=Select_poi_from (Cal_set);
ifpoi-̡th then
add_poi_to (R_Set);
Nei4=Build_neighbor_of (poi);
add_Nei_to (tmp_set);
Delete_poi_from (Cal_set);
else
add_poi_to (final_set);
Delete_poi_from (Cal_set);
}
}
END BEGIN
Ұд,׼Ϊ:Ұ4ص3,򽫸Ұںϵ,Ӷ֤,ڿն.㷨,ֹͣ,ȡͳ,ƽҶȺͻҶȾ,,.Уݸ㷨õͼ5ʾ.ָĵ2׶Ѱģ.ɿԤȻָĿĴԱ߽,ͼ5ʾ.ȻɴԵν,ֱõյıԵ.ǰ,ܹݳʼ·õһϵص㼯,ʹЩصڳʼ·,õָ꺯ֵ,ЩصжΪָıԵص..ЦΪָĳʼ·ص㼯,Nei()Ϊʼ·ĸص8صļ.,ûָ㼯O_Set,㼯C_Set,C_SetеĵΪָ꺯õص.㷨:
While O_Set NOT Empty Do begin
x=Select_one_point_of (O_Set);
Nx=Build_neighbor_of (x);
For yNx Do
Ty=R=ii*ci*(fi(y));
P=(TxTy)?Tx:Ty
add_to_C_Set(p);
Remove_P_from(O_Set);
End For
End While
ʵ,ѡôMRI豸ϻȡĹŲMRIͼ.ͼ6Ϊеһϲ,


ͼ5 ٽͼ6 ԭʼMRIͼ
ݵĳ1ʾ. 
1

ͼԴͼ߶ͼȳλϲȳˮƽֱʳֱֱʳϲ
MRI127 pixel181 pixel16bits/pixel1.700 0mm0.546 875mm0.546 875mm27

ͼ6ʾ,ָָĿ.,ڷָѡȡʼĵ,ڸõ,õͼ5ʾԤָ.õĸƽһҶֵΪ0.541 667,ƽҶȾΪ0.164 583.ıԵɴõ,ͼ5ʾ.Էָб궨,õͼ7Ľ.ͼ8Ǹָõ.


ͼ7 ʵʷָ ͼ8 ָ 
2  
Ǹ˻˼6˻ڴĽʽͼָģ,ҽ˳о.ʵ,һҽѧͼķָܹõϺõĽ,һֵ̽ͼֶָ.,кֵܶýһо,:ָ꺯ѡתѡѡȡԼֶμģ͵ĸĽȵ.,ھʵֹ,οף7ݵĲԴ,ڴ˱ʾл.
оõȻѧ͹863߿ƼĿ.,1970,ʿ,ҪоΪͼ,ģʽʶ.,1960,ʿ,оԱʿʦ,ҪоΪӾ,ͼѧ,ýϵͳ.
ͨѶϵ:ݣ 100080,йѧԺԶо
ߵλйѧԺԶо 100080
ο
1Udupa J K, Herman G T. 3D Imaging in Medicine. CRC Press, 1991
2Mortensen E, Morse B, Barrett W et al. Adaptive boundary detection using Live-Wire two-dimensional dynamic programming. In: IEEE Processing of Computers in Cardiology. New York: IEEE Press, 1992. 635638
3Kass M, Witkin A, Terzopoulus D. Snakes: active contour models. Computer Vision, 1987,1(4):321331
4Geiger D, Gupta A. Dynamic programming for detecting, tracking, and matching deformable contours. IEEE Transactions on PAMI, 1995,17(3):294302
5Rhodes M L. Computer graphics and medicine: a complex partnership. IEEE Computer Graphics and Applications, 1997,(1):2228
6ݴΪ,,.ϵͳۺϼ.:㽭ѧ,1994
(Dai Ru-wei, Wang Jue, Tian Jie. Metasynthesis of Intelligent System. Hangzhou: Zhejiang Science and Technology Publishing House, 1994)
7ݵ.ʵͼ봦.:ӹҵ,1995
(Tian Jie et al. Practical Image Processing and Analyzing Techniques. Beijing: Electronics Industry Publishing House, 1995)
1998-03-11ո壩
