Transient streamlines: texture synthesis for in vivo flow visualisation

Magnetic resonance (MR) imaging is a versatile technique for providing detailed information on blood vessel morphology and function. With its ability to acquire multi-dimensional cine flow data, MR is also an important tool for providing insight into blood flow patterns in vivo. The purpose of this paper is to describe the application of texture synthesis for flow visualisation. Two related issues are addressed, one is the removal of image noise from the acquired velocity data to ensure a correct representation of the underlying flow structure, and the other is the formation of transient streamlines through flow texture synthesis. The process of noise removal is achieved by using a convex projection algorithm based on the principle of mass conservation, whereas transient streamlines are formed via an iterative orientated pattern formation and enhancement procedure. The method described provides realistic visualisation of the flow patterns and avoids distortions caused by integration errors associated with conventional streamline tracking techniques. Effectiveness of the method applied to MR flow data acquired in healthy volunteers and patients is demonstrated.