Recurrent Nasal Papilloma Detection UsingaFuzzyAlgorithm Learning Vector Quantization Neural Network
Chuan‐Yu Chang · 2006
The objective ofthis paperistodevelop acomplete solution forrecurrent nasalpapilloma (RNP)detection. Recently, the Gadolinium-enhanced dynamicmagnetic resonance image(MRI)hasbeendeveloped andwidely usedin clinical diagnosis ofrecurrent nasalpapilloma. Owingtothe response ofRNP regions inGadolinium-enhanced magnetic resonance images isdifferent fromtheresponse ofnormal tissues, thedifference between thedynamic-MR images before andafter administering contrast material canbeusedtoextract thecoarse RNPregions automatically. Then, afuzzy algorithm forlearning vector quantization (FALVQ)neural network is usedtopick thesuspicious RNPregions. Finally, afeature-based region growing methodisapplied torecover thecomplete RNP regions. Theexperimental results showthattheproposed methodcandetect RNP regions automatically, correctly and fast. I.INTRODUCTION Since theenvironment pollution, theportion ofhumaninAsia suffers fromnasalpapilloma isincreasing gradually. Therefore, itisimportant toobtain better detection results for surgery. Recently, a new imagemodality called Gadolinium-enhanced dynamic magnetic resonance images