Fiber tractography is still unique in providing detailed imaging of white matter fiber bundles and connectivity between different brain regions. For finding specific fiber bundles, the most applied technique is tracking fibers from the seeds in a region of interest (ROI) within a diffusion tensor imaging (DTI) volume, or the limitation of tracking results to the ROI. Color-encoded fractional anisotropy (FA) map derived from DTI data, neuroanatomical atlas, and anatomical T1-weighted magnetic resonance imaging (MRI) data have been proposed as complementary data to improve the placement of an ROI. Mental mapping of colors in color-encoded FA map to directions requires a cognitive process. This paper addresses the fusion of shape with color to make the ROI drawing more a perceptual rather than a cognitive task. We propose the rendering of diffusion tensors as superquadric glyphs (shape) superimposed over the standard practice consisting of a color-encoded FA map (color) co-registered to a T1-weighted MRI image (anatomical constraint). A novel object-space algorithm that can efficiently render diffusion tensor glyphs is presented. A strategy for distributing the GPU hardware workload was devised to maximize its occupancy and reduce its stall. Implementations with a compute shader, and a geometry shader are detailed comparatively. We show that our proposal outperforms other rendering solutions. Preliminary quantitative comparisons of the nerve fibers reconstructed by interactive seeding strategies with and without the glyphs suggest that the first approach is more accurate in conveying directional information.

纤维束摄影术在提供白质纤维束的详细成像以及不同大脑区域之间的连通性方面仍然是独一无二的。为了找到特定的纤维束，最常用的技术是从扩散张量成像（DTI）体积内感兴趣区域（ROI）中的种子跟踪纤维，或者将跟踪结果限制在ROI。已经提出了从DTI数据，神经解剖图谱和解剖T1加权磁共振成像（MRI）数据派生的颜色编码的分数各向异性（FA）图作为补充数据，以改善ROI的位置。颜色编码的FA映射中颜色对方向的心理映射需要认知过程。本文着眼于形状与颜色的融合，以使ROI绘图更具感知性，而不是认知性任务。我们建议将扩散张量呈现为超二次字形（形状），并叠加在由标准配准到T1加权MRI图像（解剖学约束）的颜色编码FA映射（颜色）组成的标准做法上。提出了一种可以有效地绘制扩散张量字形的新对象空间算法。设计了一种分配GPU硬件工作负载的策略，以最大程度地提高其占用率并减少停顿。比较详细地介绍了计算着色器和几何着色器的实现。我们表明，我们的提案胜过其他渲染解决方案。通过交互式播种策略（带有和不带有字形）重建的神经纤维的初步定量比较表明，第一种方法在传达方向信息方面更准确。