肺4D-CT时空高分辨率图像构建关键技术研究

基于4D-CT的精确放射治疗已成为肺肿瘤放射治疗的重要方向。当前，肺4D-CT图像的时空分辨率还十分有限，为了提高肺4D-CT图像的时空分辨率，本项目拟研究肺4D-CT时空高分辨率图像构建新方法：（1）增加肺4D-CT图像空间分辨率的超分辨率重建策略。主要包括：融合退化算子和低秩矩阵重建的肺4D-CT超分辨率重建模型构建及求解；基于跨层面联合字典训练及多尺度稀疏表达的超分辨率重建算法。（2）增加图像时间分辨率的肺4D-CT中间相位重建策略。主要研究基于稀疏低秩分解和基于学习预测的肺图像配准算法。在精确配准的基础上重建中间相位肺图像，以增加肺4D-CT图像时间分辨率。研究成果将有望在不增加任何扫描工作和剂量耐受的条件下，构建高时空分辨率肺4D-CT图像，进而为实现放射治疗过程中的肺肿瘤4D精确跟踪奠定基础。

Precise radiotherapy based on 4D-CT plays an important role in lung cancer radiotherapy. At present, spatial and temporal resolution of 4D-CT images is still very limited. In order to improve the spatial-temporal resolution of lung 4D-CT data, this research focuses on constructing high resolution (HR) lung 4D-CT images. The major topics include: (1) Super resolution (SR) reconstruction of lung 4D-CT to increase the spatial resolution. It contains SR reconstruction based on degradation operator and low rank matrix, and sparse representation with multi-scale dictionary learning for SR reconstruction. (2) Middle phase images reconstruction to increase temporal resolution. This topic includes multi-phase lung image registration based on sparse low rank decomposition and learning-based prediction for lung image registration. Then the middle phases can be reconstructed based on the registration deformation fields. The work may effectively enhance the spatial-temporal resolution of lung 4D-CT data without increasing the scanning dose and time, which may facilitate accurate treatment planning and delivery of radiation therapy.