ارتقاء عروق كرونري در تصاوير آنژيوگرافي قلب با استفاده از بانك فيلتر جهتي

چكيده انگليسي :

Coronary Vessel Enhancement in Angiograms Using Directional Filter Bank Morteza Jalalat Vakilkandi m jalalat evakilkandi@ec iut ac ir Date of Submission 2011 10 04 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 Heart diseases are one of the rife causes of death in the world Vessel stenose is a septum disease ofvessels which happens in aorta jugular vein coronary and other vessels X ray angiography has been usedas a gold standard technique to diagnose coronary vessel stenose In angiograms there are structures andtissues therewith vessels in the background Though a technique called Digital Subtraction Angiography DSA is used to eliminate extra structures but cannot do it perfectly Therefore some algorithms areneeded to enhance vessels and remove non vessel elements in the background automatically So automaticsegmentation has been one of the major problems in clinical procedures and there have been manychallenges Blur and noisy images with quasi Gaussian vessel intensity profile and presence of extrastructure cause to failure in automatic segmentation In angiograms a distinct vessel boundry does not exist so a clear definition of accuracy is difficult In this thesis we begin to review and implement few recentefficient algorithms One of these is vessel enhancement based on smoothed directional filter bank DDFB This algorithm is robust to noise and prevents junction suppression Junction suppression is seen in Hessianbased methods Second algorithm uses directional Gabor filter bank to enhance the vessel tree Thirdalgorithm which investigated in this thesis is vessel extraction with extra edge indicator factor Thisalgorithm only appends edge indicator to vesselness filter then according to filter output determinesoptimum parameters and scale The DDFB based algorithm because of superior performance compared toother algorithms is investigated in more detail Solutions are proposed to improve this algorithm Moreover an algorithm based on smoothed fast directional filter bank FDFB is proposed which decomposes imageinto directional images and then introduce two hyper parameters to adjust filter sensitivity locally Thesehyper parameters are determined using local contrast in original image Fuzzy inference system is used toset hyper parameters in each pixel The experimental results are evaluated by two measures ROC curvesand Hausdorff distance measure HDM illustrate the superior performance of the proposed algorithm Sensitivity of the proposed algorithm is improved 8 19 and 13 2 compared to DDFB based and edgeindicator methods respectively Also specificity is improved 7 72 and 13 67 respectively Evaluatingwith HDM results in 7 24 pixels for the proposed algorithm while 15 04 and 22 65 pixels for the DDFB based and the edge indicator method respectively Vessel diameter profile contains worthy informationabout stenosis point in vessel tree So in this thesis we proposed a new algorithm to extract and plot thevessel diameter profile The result profile is smoothed by non linear diffusion filter for smoothing the curvebecause of quantization error Considering severe fall in profile this algorithm can estimate position ofstenosis in the vessel tree and then declare it in the image Keywords Digital Subtraction Angiography Vessel enhancement and extraction Directional filterbank Directional decomposition Vessel diameter profile Stenosis position estimation

جلالت وكيل كندي، مرتضي

ارتقاء عروق كرونري در تصاوير آنژيوگرافي قلب با استفاده از بانك فيلتر جهتي

تجزيه جهتي تصوير , پروفايل قطر رگ , تخمين محل گرفتگي