摘要：

脊柱的三维模型在治疗脊柱侧弯等脊柱疾病时发挥着重要的作用, 但传统的脊椎三维重建方法存在耗时长、主观性强、辐射大等问题。为应对这些挑战, 提出一种基于双平面X线片的脊柱三维重建网络X2S-Net。利用患者的正位和左侧位X线片作为输入, 通过双视角平行编码器、三维重建模块以及分割监督模块后重建出对应位置的脊柱体素模型, 实现了从X线片到可视化三维模型的端到端生成。X2S-Net在特征提取阶段, 使用了针对双平面X线片特点而设计的平行特征编码器, 用于提取脊柱的空间信息, 并设计多尺度通道注意力机制用于提取特征。在三维模型阶段, X2S-Net结合传统图像分割任务设计了分割监督模块以提高三维重建效果。实验结果表明, X2S-Net能够充分利用双平面X线片的输入信息对脊柱进行三维重建, 各数据集的平均Hausdorff距离达到了6.95 mm, Dice系数达到了92.01%。

关键词: 脊柱侧弯, 三维重建, 深度学习, 注意力机制, 计算机断层扫描

Abstract:

Three-dimensional models of the spine play an important role in the treatment of spinal disorders such as scoliosis. However, traditional methods for spinal three-dimensional reconstruction suffer from issues such as long processing times, subjectivity, and high radiation exposure. We propose a spinal three-dimensional reconstruction network based on biplanar X-ray images called X2S-Net. The network takes the anteroposterior and lateral X-ray images of the patient as input and reconstructs the corresponding voxel model of the spine using a parallel encoder, three-dimensional reconstruction module, and segmentation supervision module, achieving end-to-end generation from X-ray images to visualize three-dimensional models. In the feature extraction stage, X2S-Net employs a parallel feature encoder designed for the characteristics of biplanar X-ray images to extract spatial information of the spine and incorporates a multiscale channel attention mechanism for feature extraction. In the three-dimensional modeling stage, X2S-Net combines traditional image segmentation tasks with a segmentation supervision module to improve the three-dimensional reconstruction results. The experimental results demonstrate that this method effectively utilizes input information from biplanar X-ray images for three-dimensional reconstruction of the spine, achieving an average Hausdorff distance of 6.95 mm and a Dice coefficient of 92.01% across the datasets.

Key words: scoliosis, three-dimensional reconstruction, deep learning, attention mechanism, Computed Tomography(CT)