6766

6308

كارشناسي ارشد

كنترل

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1390

[ده]، [106]ص.: مصور، جدول، نمودار

ص.ع.به فارسي و انگليسي

مريم ذكري

نيلوفر قيصري

28/3/91

رسول امير فتاحي، حسين رباني

1396/09/14

كتابنامه

برق و كامپيوتر

مهندسي برق و كامپيوتر

ID6308

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

109 Malignant Melanoma Diagnosis from Dermoscopy Images Using Wavelet Networks Amir Reza Sadri ar sadry@yahoo com February 21 2012 Department of Electrical and Computer Engineering Isfahan University of Technology 84156 83111 Iran Degree M Sc Language Farsi Supervisor Dr Maryam Zekri m zekri@cc iut ac ir Abstract The most dangerous type of skin cancer is Malignant Melanoma MM which is caused by irregular development of skin pigment At present this cancer is increasingly growing with a rate of 50 Nevertheless it is also the most treatable kind of skin cancer if diagnosed at an early stage The most certain way of diagnosing MM is biopsy Biopsy which is an attacking approach of diagnosis has major pitfalls Furthermore since the appearance of MM is similar to other skin lesion such as spots in many cases there is no need to undergo the harsh process of sampling Thus using non invasive approaches are better choices for diagnosing the MM among which the best technique is Dermoscopy Imaging DI In DI a dermatoscope is used to take images from the lesion then these images are analyzed by a dermatologist Although dermoscopy images have a great capability of immediate diagnosis of MM their interpretation is quite time consuming and demanding on the part of specialists Therefore recently there has been a tremendous tendency towards using computer and intelligent devices for interpreting and analyzing the dermoscopy images and there is a great interest in the development of Computer Aided Diagnosis CAD systems that can assist the clinical evaluation of dermatologists The present study employs Wavelet Network WN as an intelligent system This WN is a member of Fixed Grid Wavelet Network FGWN that is formed with no need of training The network structure after normalization of input is formed with selecting the mother wavelet shifts and the scales coefficients of wavelet lattice Then through a stage of selection the effective wavelets are selected with Orthogonal Least Square OLS algorithm This WN is used in three stages of noise cancellation segmentation and diagnosis of the disease The noise which is usually produced on dermoscopy images is of the impulse type and the proposed algorithm shows proper results for its cancellation After noise removal which is a pre processing stage segmentation and bordering of the lesion are done with the aid of the proposed WN In order to evaluate the results of the proposed method they are compared with the results obtained by a pathologist In the next stage considering the Asymmetry Border Color Diameter ABCD rule some features of the lesion whose border was determined before are extracted Then in order to reduce the measurements time and increase the precision of calculation the number of features are reduced This stage of work is done by Principle Component Analysis PCA Sequential Forward Selection SFS and under the supervision of a dermatologist The best result is obtained by SFS Finally malignant melanoma is diagnosed and classified by proposed WN as a classifier The specificity and sensitivity of the algorithm is 93 and 94 respectively It should be mentioned that the images used in this study are taken from a valid database and this thesis is carried out in line with a research project for Medical Image and Signal Processing Research Center Isfahan University of Medical Sciences Keywords Wavelet Network Wavelet Lattice Orthogonal Least Square Segmentation Feature Selection Classifier

مريم ذكري

نيلوفر قيصري

رسول امير فتاحي، حسين رباني