Automated segmentation and classification of cervical cells using deep auto encoder based extreme learning machine

Abstract

Cervical cancer is one of the deadliest cancers known and is also a key research area in image processing. Cervical cancer is curable if it is detected timely and treated appropriately but detecting it at the pre-cancerous stage remains a challenging task. Screening of cervical cancer is a major problem since most cervical cancer screening procedures are invasive in nature. Hence, there is hesitancy for normal screening procedures and also more time is required for knowing the results. Although traditional screening technique like the pap-smear test has been proved highly successful in screening individuals, the main drawback of this technique is the complexity and the problems involved in the preparation of smear. The pap-smear test involves a great deal of skill that is associated with the screening procedure. It can be done only by a trained and knowledgeable cytotechnologist or a pathologist. newlineBut the proposed method of screening using the pap-smear imaging technique is quite simple and fast compared to the traditional methods of screening cervical cancer. This method can also be deployed for a large number of cases quite fast and accurately. Hence this proposed research evolves a technique which is involving a pap-smear image of the cervix tumor region. It presents a digital imaging system able to assist physicians to track cervical cancer. The goal is to automatically extract the region where cervical cancer starts to occur. Pap-smear imaging techniques are one of the tools to diagnose cancer and identify the malignant, benign tissue in the human body. newline newline