Creating a Optimal Set of Textural Features for Cervical Cancer Lesions Using Hierarchal Clustering Technique

Abstract—Cervical cancer is considered to be the deadliest gynaecological cancers which affect women and it now seems that the only way it could be controlled is by employing an automated system. Of all the visual parameters which the brain uses to make sense of the world around us the textural features is the best reliable one. Here we have tried to exploit this concept by analyzing the textural features of cervical cyto images. The Nucleus of a cervical cyto image has with it varied textural information which inherently tell about the stage of the cancer. There are many textural features available and since each image has different properties no two images of the same stages will show positive for a particular feature. Hence the alternative is to have combination of features. But in doing so you must have a right balance between speed and accuracy of the system. Hence in this work apart from extracting the features we have proposed a ranking system which will help us to create a balanced set of combinational features so that it won’t affect the accuracy nor the speed of the automated system