Semantic segmentation of angiographic images

Thc ovcrall purpose of this work is to identify objects in an angiographic sequence by exploiting the tem- poral correlation between adjacent frames for analysis and compression purposes. The detection of the vascular tree in a reference image can support segmentation in adjacent frames by reducing the detection problem to a tracking pro- cedure along the sequence. Object identification also allows an object-oriented approach to compression, as suggested by medical images nature. A vessel segmentation algorithm has been designed and implemented; it is an extrapolation and update procedure which exploits the a priori knowledge about the image content.