Although digital images indexing and querying techniques have extensively been studied for the last years, few systems are dedicated to medical images today while the need for content-based analysis and retrieval tools increases with the growth of digital medical image databases. We analyze medical image properties and we evaluate Gabor-filter based features extraction for medical images indexing and classification. The goal is to perform clinically relevant queries on large image databases that do not require user supervision. We demonstrate on the concrete case of cardiac imaging that these techniques can be used for indexing, retrieval by similarity queries, and to some extend, extracting clinically relevant information out of the images.