Separation of perfusion phases in angiographies

The analysis of Cerebral Angiographies are an essential tool for the assessment of the future of patients that underwent thrombolysis after a stroke event. Many semi-qualitative visual diagnostic scales have been developed for this purpose. Perfusion angiographies show essentially three phases: the arterial (early), the capillary (intermediate), and venous (late) phase. We call parenchymogram the image sequence corresponding to the capillary phase only. Unfortunately the parenchymogram is often under exploited in practice, despite containing many pertinent hints on the quality of reperfusion. In this paper we propose a set of methods for the extraction of the parenchymogram from raw Cerebral Angiographies. These methods rely on basis pursuit and on the representation of images with an over-complete basis arising from an redundant wavelet transform. We will show that the extraction of the parenchymogram by applying the aforementioned methods on real clinical data allows us to recover essential information for the comparison of blood flow before and after thrombolysis.