Image processing in echography and MRI
This work deals with image processing for three medical imaging applications: speckle detection in 3D ultrasound, left ventricle detection in cardiac magnetic resonance imaging (MRI) and flow feature visualisation in velocity MRI. For speckle detection, a learning from data approach was taken using pattern recognition principles and low-level image features, including signal-to-noise ratio, co-occurrence matrix, asymmetric second moment, homodyned k-distribution and a proposed specklet detector. For left ventricle detection, template matching was used. Forvortex detection, a data processing framework is presented that consists of three main steps: restoration, abstraction and tracking. This thesis addresses the first two problems, implementing restoration with a total variation first order Lagrangian method, and abstraction with clustering and local linear expansion.