Interaction Model for Large Radiologic Image Collections

Thesis Work, awarded. Visualization and interaction with large quantity of images.

The main contribution of this thesis is a framework to improve the interaction with large radiological image collections, integrating in a single environment diverse computational strategies. Starting with the characterization of brain, breast and spine magnetic resonance images. Using different image properties like: texture, intensity and shape. Subsequently, a classification mechanism is used to automatically identify the organ to which the image belongs to. A content-based image retrieval strategy is used to compute the distance among cases, later these are used to graph a visual resume of the available volumes in the database. Using this visual summary or searching for an specific case, the user may study on detail each volume in the repository.