Computer-aided Diagnosis Technologies in Medicine

In this thesis, I focused on the research stream of computer-aided diagnosis technologies in medicine, and together with my collaborators I proposed several approaches for certain applications. Chapter 2 was motivated by the need of the exclusive detection of vascular bifurcations in retinal images.I demonstrated the effectiveness of the proposed model in two applications. One application concerns the detection of architectural and electrical symbols and the other one is the exclusive detection of vascular bifurcations without crossovers in retinal fundus images. In Chapter 3, Chapter 4 and Chapter 5, I proposed methods that can be used to assist medical experts in the diagnosis of epidermolysis bullosa acquisita (EBA). In Chapter 3, I reported a modified inhibition-augmented model for the ridge-ending detection, which is used for localizing u-serrated patterns for the diagnosis of EBA. In Chapter 4, I gave an account of another novel approach of automatic differentiation of u- and n-serrated patterns by normalized histogram of orientations in DIF images. In Chapter 5, I investigated the feasibility of using CNNs for the recognition of u-serrated patterns that can assist in the diagnosis of EBA.