Designing a secure ubiquitous mammography consultation system

Abstract

This thesis attempts to design and develop a prototype for mammography image consultation that can work securely within a ubiquitous environment. Mammogram images differ largely from other type of images and it requires special and dedicated techniques to identify the required regions of interest. Thus in Chapter 2 we started to explore the affectivity of the various traditional techniques based on convolution operators (e.g. Sobol, Pretwitt, Canny) for mammography edge detection. The second part of chapter 2 tries to enhance the results obtained via the traditional techniques by hybriding some of them. The hybriding technique is called in our thesis as Pipelined Operators. In this direction we proposed four pipeline operators, which contribute to the edge enhancement as well as abnormalities rendering through the introduction of an additional coloring mechanism. Although the visualization pipelines represent in our view an advancement on the traditional techniques applied to mammograms, such pipelines expose healthcare users to further usage complexities. For this purpose we extended our research work in chapter 2 to find a better single technique that can work smoothly within the healthcare system. In this direction, we developed in the third part of chapter 2 a novel technique for finding edges based on analyzing the dynamic and fuzzy nature of edges in mammograms. We called our developed method as "Dynamic Fuzzy Classifier or the DFC".