RESUMO
Recently, extensions to curved planar reformation (CPR) were proposed to improve vascular visualization of medical images. While these projective transformations provide enhanced visualization of vascular trees, non-planar alignment and arbitrary topology can cause visualization artifacts. Vascular trees in medical images are not aligned to planar cross-sections of volumetric image slices and thus aggravate simultaneous visualization of diagnostic features. Complex tree topology and non-planar alignment requires the need for an adaptive projection scheme to prevent visualization artifacts while preserving correctness of anatomical information. In this paper, we present algorithmic details for topological and orientation invariant visualization of vascular trees. Vascular high-level description of the medial axis guides the reformation process by flattening the vascular tree interior to successive image planes for respective radial sampling angles. Tree orientations are estimated from intrinsic shape properties of the vascular tree for rotation invariant projection. Radial sampling planes perpendicular to the medial axis tangents are the basis for topological invariant visualization of complete vascular interiors. We present experimental results on two different vascular tree topologies and demonstrate that our method is able to produce artifact free visualization of vascular interiors.