A Study on Automatic Detection of Uterine' Cervical Pap- Smears by Image Processing

Cancer of the cervix is the most common malignancy in women in developing countries and the second most common cancer in women throughout the world with approximately 500,000 new cases each year. Prevention of this large number of premature deaths among women is, therefore, a goal worthy of urgent and serious consideration. Due to its high diagnostic disagreement among pathologists and large quantity of specimens, it is necessary to develop an automatic screening system measuring morphologic and densitometric features of the samples. Many research works have been published but most of them used Feulgen stained specimens which are not a usual staining method used in clinics. In this thesis, an automatic cancerous nucleus detection method essential to a screening system with papanicolaou stained specimens called Pap-smear is proposed which employs image processing techniques. It uses edge information to segment objects and morphologic as well as densitometric information to distinguish cancerous nuclei from dirts or normal nuclei. It has produced useful results in our study.