Abstract: Over the last years significant improvements in image processing for ophthalmology were introduced. Retinal imaging is an important modality to document both the health of human eyes and their biometric features. Recent advances in automated diagnostic systems actually enable eye doctors to perform a large number of screening exams for frequent diseases, such as diabetes or glaucoma. On this proposal, precise information about the Optic Disk (OD) reveals necessary to examine the severity of some diseases, since changes in the OD often indicate a pathologic progression. Diagnostic systems of analysis are usually based on the detection of the so-called Region of Interest (ROI) in the optic disc, which is a subset of the image domain important for each retinal analysis. In detail, a properly extracted ROI provides a smaller image containing the most diagnostic information and much less time-consuming when processed. Thus, in retinal imaging the localization of the OD, in terms of position of its centre and length of radius, has to be the best possible, being OD viewed as the main reference when analyzing every anatomic/pathologic retinal detail and the detection of its centre C as a key step in automatic extraction of retinal anatomic features. Since errors caused by artifacts and bright fringes on retinal images could affect results in image processing, in this speech a proposal for an accurate identification of the centre of OD is presented, which combines more systems to analyze retinal images in parallel and find possible values of the centre of optic disc in different ways. Locations of resulting pixels can then be processed to identify the point to adopt as the centre of the optic disc as effectively as possible.

Brief Biography of the Speaker:
Leonarda Carnimeo is an Associate Professor of Electrical Sciences at the Engineering Faculty of the Politecnico di Bari, Bari, Italy. Since 1988 her scientific activity dealt with various aspects in the field of Circuit Theory and Applications. At first, her research interests concerned with various aspects of analysis and decoupling of electric power networks, stability analysis of continuous-time and discrete-time cellular neural networks, associative memories design, synchronization of chaotic circuits with applications to secure communications. She was successively involved in various research activities concerning with the development of new techniques for the design of NN-based/CNN-based associative memories with application to artificial vision systems. Then, she was the Responsible in a 2001 Italian PRIN project in the field of Neural Networks and Nonlinear Circuits. Later she was the Responsible in an Italian PRIN project 2004-2006 for the Synthesis of a Cellular Neural System and Development of Innovative Techniques for a Real-time Ophthalmic Image Processing System as a Support in Retinal Pathology Diagnosis and Laser Therapy. She is author of more than 100 scientific papers, with a significant amount published on International journals and chapters of books. Prof.Leonarda Carnimeo, member of the National Group of Coordination of Electrical Sciences, develops her scientific activity within the following interdisciplinary themes: Synthesis of neural systems for retinal imaging in aid of diagnostics and biometric identification; Nonlinear cellular architectures for the spatial novelty detection in stereoscopic vision systems; Cellular associative memories for statistical process control; Modelling of the european Interchange Energy Network.