Effective FuzzyClustering Algorithm forAbnormal MR BrainImage Segmentation

Clustering approach iswidely usedin interest recently intheuseoffuzzyclustering biomedical applications particularly forbrain methods, whichretain moreinformation fromthe tumordetection inabnormal magnetic resonanceoriginal imagethanhardclustering methods. Fuzzy (MR)images. Fuzzyclustering usingfuzzyC- C-means algorithm iswidely preferred because ofits means(FCM)algorithm provedtobesuperioradditional flexibility whichallows pixels tobelong to overtheotherclustering approaches intermsof multiple classes withvarying degrees ofmembership. segmentation efficiency. Butthemajordrawback Butthemajor operational complaint isthat theFCM oftheFCM algorithm isthehugecomputational technique is timeconsuming(3).Several timerequired forconvergence. Theeffectiveness modifications havebeendoneontheexisting network oftheFCM algorithm intermsofcomputational toimprove theperformance. rateisimproved bymodifying thecluster center A hierarchical FCM algorithm basedontemplate andmembership valueupdation criterion. Inthis matching isproposed byKwon& Han(4). Butit paper, theapplication ofmodified FCM algorithmsuffers fromthedrawback oftherequirement foran forMR braintumordetection isexplored.accurate template. Thesegmentation efficiency ofthe AbnormalbrainimagesfromfourtumorclassesFCM algorithm isimproved bysilhouette method namelymetastase, meningioma, gliomaand basedcluster center initialization instead ofrandom astrocytoma are used in thiswork. A initialization (5). TheFCM algorithm hasbeenalso comprehensive feature vector spaceisusedforthe implemented using theconcept ofparallel processing segmentation technique. Comparative analysis in (6). Eventhough itpromises highspeed processing, termsofsegmentation efficiency andconvergencethehardware implementation isnoteffective. Cheng rateisperformed between theconventional FCM andGoldgof(7)proposed thefastclustering andthemodified FCM.Experimental results show algorithm based onrandomsampling whichyields a superior results forthemodified FCM algorithmspeed-up factor of2-3times whencompared withthe intermsoftheperformance measures. conventional FCM algorithm. Thevector quantization basedFCM algorithm hasbeenimplemented anda Indexterms -Clustering, MR braintumor,Fuzzy nominal speed-up factor isachieved (8). Fastfuzzy C-means andSegmentation efficiency.