INTRODUCTION Due to the multi-dimensionality of the data, visualization of diffusion tensor fields, hence tissue microstructures, has generally required a tradeoff between the information included and spatial resolution of the representation. For example, although object-based renderings (e.g., ellipsoids [1]) can simultaneously depict the diffusivities and orientations of all principal diffusion axes, spatial resolution and continuity are sacrificed to avoid visual cluttering. In contrast, pixel value-based representations (e.g., falsecolor-coded fiber angles [2]) circumvent these spatial issues, but only partial information regarding diffusion anisotropy is shown each time, and the viewer often needs to be re-trained to correctly interpret the coding schemes. In this study, we applied a line-integral convolution (LIC) technique [3] to visualize diffusion tensor fields obtained via DTI [4]. Special considerations were incorporated to extend the original 2D method for rendering 3D diffusion tensor fields, and to reconstruct both the principal and second-order anisotropy of water diffusion. The generalized LIC technique is demonstrated in reconstructing myocardial fiber and lamellar architectures. Results indicate the approach to be robust, scalable, and share key advantages of both objectand pixel value-based means to visualize tissue microstructure.