DTI volume rendering techniques for visualising the brain anatomy

Abstract

Over the past few years Diffusion Tensor Imaging (DTI) has become an increasingly popular method for imaging the brain anatomy and diagnosing a variety of neurodegenerative diseases. Unfortunately the size and multi-dimensional nature of diffusion tensor data sets makes it difficult to understand them. We use illuminated streamlines to compute high quality dense 3D visualisations of the 3D nerve fibre structure. Nerve fibres are extracted using a numerical integration technique and a fuzzy classifier which represents the probability that a sample point represents grey matter, white matter or Cerebral Spinal Fluid (CSF). We present two novel methods which improve the perception of the 3D arrangements of fibre tracts. The first method is a hardware accelerated algorithm which represents fibres as semi-transparent tubes with emphasised silhouettes. Because of the semi-transparent nature of the tubes inside structures are revealed. The enhancement of tube silhouettes improves the identification of individual fibre tracts and their 3D arrangement. The second method uses direct volume rendering and multiple colour and transparency look-up tables to represent the directional information of the nerve fibre structure and other tissue types simultaneously. The method can be used to represent finer details depending on the resolution of the noise texture employed. Depending on the choice of the opacity transfer functions fibre tracts can be represented semi-transparent or nearly opaque.

Ray Tracing Generalized Tube Primitives: Method and Applications

“…Such line primitives are used to represent data in a range of scientific domains, such as fluid dynamics (e.g., streamlines and pathlines) [Ste00, MTHG03, STH*09, GGTH07, Mer12], medical imaging (e.g., diffusion tensor imaging) [RBE*06, MSE*06, ZDL03], and vector field visualization (e.g., magnetic or vector fields) [PVH*02, CYY*11, MCHM10]. Additional attributes can be encoded along the line by varying the line color, thickness [LMSC11], or opacity [WVDLH05, GRT13, KRW18]. This same type of geometry—long, thin lines with varying thickness—is also useful for representing other data, such as ganglions in neuron datasets [Mar06] or vessels in aneurysm visualization [SSV*14], although such data further requires the method to support acyclic graph structures.…”

Fiber Visualization with LIC Maps Using Multidirectional Anisotropic Glyph Samples

“…By modifying different parameters of the input image, Hotz et al [ 22 ] encoded the eigenvalues of a positive-definite metric with the same topological structure as the tensor field. Wunsche and Linden [ 23 ] proposed a three-dimensional LIC volume to be visualized by direct volume rendering. It is also possible to compute a single LIC image for each of the three eigenvectors and overlay the resulting images to get a fabric-like texture [ 22 ].…”