Detection of Specular Reflection and Segmentation of Cervix Region in Uterine Cervix Images for Cervical Cancer Screening

Abstract Background: Visual Inspection with acetic acid is a screening method for detecting cervical cancer in resource poor settings. Pre-cancerous and cancerous regions turn white on combining with acetic acid. They are called acetowhite regions and can be considered as the indicators of abnormality. Specular reflections, which are bright white regions, interfere with the detection of acetowhite regions and hence need to be eliminated. The irrelevant regions in the cervix images such as medical instruments, vaginal walls etc., need to be eliminated for better processing efficiency. Methods: In this paper, we propose an algorithm for specular reflection detection using a standard deviation filter and cervix region segmentation using curvilinear structure enhancement. The specular reflection detection algorithm was tested on 151 cervix images. An expert compared the performance of this algorithm with manual evaluation. The cervix border detection algorithm was also tested on the same cervix image dataset. Results: ROI detection was found to have a sensitivity of 96.75% and a Dice index of 91.72%. Conclusions: The comparison of proposed method with state of the art algorithms demonstrated that the proposed method is more robust, sensitive and accurate in terms of overlapping metrics.