This paper explains general model-based approaches to two medical image pattern recognition applications. The first application is Bayesian segmentation of breast regions of interest i.e: pectoral muscle, fatty and fibroglandular regions, using Markov Random Fields. This work is a part of a computer aided diagnosis project aiming at evaluating breast cancer risk and its association with texture characteristics of regions of interest on digitized mammograms. Owing to obtained segmentation results, the proposed method could be considered as a satisfying first approach for segmenting regions of interest in a breast. Another dermatology application with the aim to classify skin images of several disorders is briefly discussed. The second part of our contribution will give an overview of the achievements of another project that has beentaking place since 2000 in the UK. The Medical Image Management and Assessment System is developed in cooperation between QinetiQ, Inc. and Universities of Cambridge, Oxford and Surrey. We will show how image analysis tools form the basis of a telemedicine system targeted at early breast cancer grade assessment. The system is capable of processing images uploaded from remote sites. It enables image analysis and classification based on combined outputs of several co-operating automated classifiers, textual and graphical annotation of processed images, semi-automatic diagnosis generation, natural language text query processing, and more. This is an example of a visionary approach that, in the long term, may lead to substantialsavings of breast cancer diagnosis cost by eliminating unnecessary thorough examination of false-alarm cases.