Segmentation of cervix using minimum spanning superpixel tree detector

Abstract

Cervical cancer occurs in cervix region of women which is life threatening and is a key research in the field of medical diagnosis. Cervical cancer can be prevented if precancerous changes is diagnosed earlier and cured properly. But finding cancer at the earlier stage is a tedious process. Cervical cancer screening is crucial since utmost of the screening procedures are invasive in nature. The objective is to automatically extract the area where in the cervical cancer starts to occur. In order to diagnose cervical cancer and distinguish the malignant and benign tissues, a cytology image obtained by cervicographic or colposcopy device is used. A robust segmentation algorithm is proposed that uses superpixel for the image segmentation method. This method is regularized by the combination of structural properties such as color, texture, and spatial data present in the superpixel graph.