ارتقاء تصاوير و جداسازي غير زمان حقيقي سلول هاي اپي تليال دهانه رحم در تصاوير تست پاپ اسمير

چكيده انگليسي :

Automatic Off Line Enhancement and Segmentation of Epithelial Cervical Cells in Pap Smear Images Negar Mohaghegh Harandi nmohaghegh@ec iut ac ir Date of Submission 06 04 2009 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisors Saeed Sadri sadri@cc iut ac ir Rassul Amirfattahi fattahi@cc iut ac ir Abstract Cervical cancer is the fifth most deadly cancer in women worldwide Human Papilloma viruses are the main risk factor for cervical cancer Due to activities of these viruses the shape and the texture of cytoplasm and nucleus of cells in cervix will change These Malignant changes can be examined in scenes of Pap smear test under Microscope by pathologists in medical centers Cervical cancer can be mostly treated if detected in early stages This procedure takes enormous time and energy from the pathologists reviewing thousands of cells each day One of the main steps to help the pathologist is to develop an automatic system which can segment and classify cells in normal and abnormal groups This project focuses on segmentation of detectable cells from background of images taken from a smear s scene by a digital camera in 10X resolution of microscope The procedure has two main steps localization of epithelial cells in low resolution images and detection of cytoplasm and nucleus contours in higher resolution In particular medical images are often corrupted by noise and sampling artifacts which can cause considerable difficulties when applying classical segmentation techniques such as edge detection and thresholding As a result these techniques either fail completely or require some kind of post processing step to remove invalid object boundaries in the segmentation results Geometric active contours have turned out to be very powerful and effective in the segmentation of medical images in recent years Also the level set formulation leads to a partial differential equation for evolution of contour which provides automatic splitting and emerging of the evolving curve In addition narrow band implementation of level set approach can reduce the consumed time by a large factor making the method suitable for near real time medical applications In this project geometric active contour without re initialization is used as a powerful tool to localize and segment the cervical cells from background of Pap smear images Also different techniques of de noising and contrast enhancement such as nonlinear diffusion filtering histogram equalization and trim meaning is examined and applied One of the main steps of the algorithm includes binarization which is performed by adapting Sauvola method for cell images This method was originally introduced for binarization of document images This project considers segmentation of connected cells as well as free lying cells in the Pap smear images In order to find the cytoplasm contour of the connected cells each of them is first modeled by a circle This procedure is performed by adapting an automatic decomposition model to the application At the end the results are presented for about 100 cells In addition the multi stage algorithm of a recently introduced paper in literature which was published in an authentic journal of image processing is implemented and its results are compared with the results of this project Also the defects of this algorithm are mentioned Key Words Cervical Cancer Pap Smear Test Image enhancement Segmentation Geometric active contours

محقق هرندي، نگار

ارتقاء تصاوير و جداسازي غير زمان حقيقي سلول هاي اپي تليال دهانه رحم در تصاوير تست پاپ اسمير

جداسازي تصاوير سلولي , كانتورهاي فعال هندسي و پارامتري , سرطان دهانه رحم