Abstract：
Diffusion-based tractography in the optic nerve requires sampling strategies assisted by anatomical landmark information (regions of interest [ROIs]). We aimed to investigate the feasibility of expert-placed, high-resolution T1-weighted ROI-data transfer onto lower spatial resolution diffusion-weighted images. Slab volumes from 20 volunteers were acquired and preprocessed including distortion bias correction and artifact reduction. Constrained spherical deconvolution was used to generate a directional diffusion information grid (fibre orientation distribution-model [FOD]). Three neuroradiologists marked landmarks on both diffusion imaging variants and structural datasets. Structural ROI information (volumetric interpolated breath-hold sequence [VIBE]) was respectively registered (linear with 6/12 degrees of freedom [DOF]) onto single-shot EPI (ss-EPI) and readout-segmented EPI (rs-EPI) volumes, respectively. All eight ROI/FOD-combinations were compared in a targeted tractography task of the optic nerve pathway. Inter-rater reliability for placed ROIs among experts was highest in VIBE images (lower confidence interval 0.84 to 0.97, mean 0.91) and lower in both ss-EPI (0.61 to 0.95, mean 0.79) and rs-EPI (0.59 to 0.86, mean 0.70). Tractography success rate based on streamline selection performance was highest in VIBE-drawn ROIs registered (6-DOF) onto rs-EPI FOD (70.0% over 5%-threshold, capped to failed ratio 39/16) followed by both 12-DOF-registered (67.5%; 41/16) and nonregistered VIBE (67.5%; 40/23). On ss-EPI FOD, VIBE-ROI-datasets obtained fewer streamlines overall with each at 55.0% above 5%-threshold and with lower capped to failed ratio (6-DOF: 35/36; 12-DOF: 34/34, nonregistered 33/36). The combination of VIBE-placed ROIs (highest inter-rater reliability) with 6-DOF registration onto rs-EPI targets (best streamline selection performance) is most suitable for white matter template generation required in group studies.
摘要：
视神经中基于扩散的纤维束造影需要解剖标志信息（感兴趣区域[ROI]）辅助的采样策略。我们的目的是研究专家安置的可行性，高分辨率T1加权ROI数据传输到较低空间分辨率的扩散加权图像。获取来自20名志愿者的板坯体积并进行预处理，包括失真偏差校正和伪影减少。约束球形反卷积用于生成定向扩散信息网格(纤维取向分布模型[FOD])。三位神经放射学家在扩散成像变体和结构数据集上标记了标志。将结构ROI信息（体积内插屏气序列[VIBE]）分别记录（与6/12自由度[DOF]成线性）到单发EPI（ss-EPI）和读出分段EPI（rs-EPI）体积上，分别。在视神经通路的靶向纤维束造影任务中比较了所有八个ROI/FOD组合。专家中放置的ROI的评估者间可靠性在VIBE图像中最高（较低的置信区间为0.84至0.97，平均0.91），在ss-EPI（0.61至0.95，平均0.79）和rs-EPI（0.59至0.86，平均0.70）中均较低。基于流线选择性能的示踪成像成功率在注册到rs-EPIFOD(6-DOF)的VIBE绘制ROI中最高(70.0%超过5%-阈值,上限与失败的比率39/16），其次是12-DOF注册（67.5％；41/16）和未注册的VIBE（67.5％；40/23）。在ss-EPIFOD上，VIBE-ROI数据集总体上获得的流线较少，每个流线均为55.0％，高于5％的阈值，并且上限与失败的比率较低（6-DOF：35/36；12-DOF：34/34，未注册33/36）。VIBE放置的ROI（最高的评分者间可靠性）与6-DOF注册到rs-EPI目标（最佳流线选择性能）的组合最适合于小组研究所需的白质模板生成。
