Most of tractography studies on insomnia disorder (ID) have reported decreased structural connectivity between cortical and subcortical structures. Tractography based on standard diffusion tensor imaging (DTI) can generate high number of false-positive streamlines connections between gray matter regions. In the present study, we employed the convex optimization modeling for microstructure informed tractography-2 (COMMIT2) to improve the accuracy of the reconstructed whole-brain connectome and filter implausible brain connections in 28 patients with ID and compared with 27 healthy controls. Then, we used NBS-predict (a prediction-based extension to the network-based statistic method) in the COMMIT2-weighted connectome. Our results revealed decreased structural connectivity between subregions of the left somatomotor, ventral attention, frontoparietal, dorsal attention and default mode networks in the insomnia group. Moreover, there is a negative correlation between sleep efficiency and structural connectivity within the left frontoparietal, visual, default mode network, limbic, dorsal attention, right dorsal attention as well as right default mode networks. By comparing with standard connectivity analysis, we showed that by removing of false-positive streamlines connections after COMMIT2 filtering, abnormal structural connectivity was reduced in patients with ID compared to controls. Our results demonstrate the importance of improving the accuracy of tractography for understanding structural connectivity networks in ID.

大多数关于失眠症（ID）的纤维束成像研究都报告了皮质和皮质下结构之间的结构连接性降低。基于标准扩散张量成像 (DTI) 的纤维束成像可以在灰质区域之间生成大量假阳性流线连接。在本研究中，我们采用微结构信息纤维束成像-2 (COMMIT2) 的凸优化模型来提高 28 名 ID 患者重建全脑连接组的准确性，并过滤掉不可信的大脑连接，并与 27 名健康对照者进行比较。然后，我们在 COMMIT2 加权连接组中使用了 NBS-predict（基于网络的统计方法的基于预测的扩展）。我们的结果显示，失眠组的左侧躯体运动、腹侧注意力、额顶叶、背侧注意力和默认模式网络的子区域之间的结构连接性降低。此外，睡眠效率与左额顶叶、视觉、默认模式网络、边缘、背侧注意力、右背侧注意力以及右默认模式网络内的结构连接之间存在负相关。通过与标准连接分析进行比较，我们发现，通过在 COMMIT2 过滤后去除假阳性流线连接，与对照组相比，ID 患者的异常结构连接减少了。我们的结果证明了提高纤维束成像的准确性对于理解 ID 中的结构连接网络的重要性。