An efficient approach to automated segmentation in medical image analysis

Automatic image segmentation is widely used in medical image analysis. However, an efficient way of automated segmentation is difficult to achieve. In this thesis, a survey of current image segmentation methods and their possible applications to identify Cervical Intraepithelial Neoplasia (CIN) are introduced. Approaches to Cervix image segmentation and analysis are discussed. A very efficient algorithm for segmentation of acetowhite regions is developed, verified and compared with other existing methodologies in this thesis. Several image processing methodologies and mathematical operations are exploited and applied to this research work. Although the success of the applied algorithms is highly dependent on the quality of the images used, statistical results regarding the feature extraction, running time and pattern classification are obtained and found to be quite satisfactory. Further identification and classification of some of the lesions within the acetowhite regions of the uterine cervix have also been achieved. This efficient automatic segmentation and classification methodology will greatly facilitate content-based image retrieval from digital archives of cervix images and has the potential of playing a significant role to the development of an image-based screening tool for Cervical Cancer.