Developing computer systems for automated diagnosis from medical images has been an area of intense research in recent years. Digital medical images are found in the vast majority of diagnostic laboratories, which allows computer systems to manage them flexibly. In particular, digital medical image processing applies techniques that extract features describing human tissue accurately and reliably (for example, segmentation). Describing biomedical images by the values of specific features allows machine learning methods (using well-known classifiers) to support automated diagnosis and assist physicians. Such systems have enormous practical value: they help detect important cases early and reliably and can be integrated both in the local information systems of diagnostic centres and in wider telemedicine systems. This thesis will explore the potential of advanced machine learning techniques such as deep neural networks.
