7013

6555

كارشناسي ارشد

مخابرات

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1390

هشت، 106ص.: مصور، جدول، نمودار

ص.ع. به فارسي و انگليسي

محمدرضا احمدزاده

نيلوفر قيصري

8/7/91

1396/09/18

كتابنامه

برق و كامپيوتر

مهندسي برق و كامپيوتر

ID6555

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Investigation of Efficient Algorithms for Sperm Detection and Tracking in Order to Classify Sperms Fertility Yoones Imani y imani@ec iut ac ir Date of Submission 2012 03 4 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSupervisor Dr Mohammad Reza Ahmadzadeh Ahmadzadeh@cc iut ac irAbstract Infertility has been widely investigated in recent years In average 40 up to 60 of all couplesexperiencing infertility a male factor is involved The first and non invasive test that checks the quantityand quality of male is assessment of semen or spermogram Since 15 years ago computer aided systems forsperm analysis CASA has been introduced The CASA system also makes the assessment process moreobjective and accurate Sperm s density mobility progressive motion and their morphology features are themost important microscopic characteristics which are used to evaluate male fertility and specifying thesperm cells health World Health Organization has determined these properties for specifying the spermcells health Sperm mobility assessment from video images is one of the most useable of these features that needaccurate tracking process However a few researches in this area have been conducted In this research wefirst study mathematical background needed to develop object recognition and tracking One of the mainsteps is the background subtraction from video images We choose the difference of consecutive frames as asimple and effective method however it suffers from selected threshold as the output depends highly onthe threshold To eliminate its dependence on thresholding we propose a nonlinear diffusion filtering intime domain Nonlinear diffusion filtering is a smoothing and noise removing method from images thatpreserves edges In tracking step we investigated many tracking methods and classified them into threecategories point tracking kernel tracking and contour tracking Then with introduction of advantages anddisadvantages of each category we assessed ability of them in our application qualitatively Based on thisassessment we choose deterministic point tracking method as our tracking method In deterministic pointtracking approach center of detected cells is assumed as the points that should be tracked We introduce anew cost function for matching of detected cells in two consecutive frames based on limitation of spermcells velocity Then a Hungarian search method is used for choosing the best matching between all possiblecandidates Experimental results show that the accuracy of one sperm tracking with two criteria MSE andframe based error is 2 62 with standard deviation 1 27 and 2 52 with standard deviation 1 74 respectively These errors are equivalent to 97 38 and 97 48 accuracy in one sperm tracking In evaluation of multisperm tracking frame based error with new definition results in 3 24 error for up to 10 sperms Keywords Infertility Spermogram Point Tracking Nonlinear Diffusion Filtering

محمدرضا احمدزاده

نيلوفر قيصري