Despite the extensive attention attracted by magnetic resonance imaging (MRI) in the radiation therapy, computed tomography was reintroduced by the researchers. During the calculation process of the 3D dose distribution of tissues, there were some arguments about the electron density information obtained from the CT scan. However, the CT-provided bones are accurate for constructing a radiograph. Recently, the advantages boosted by the soft tissue contrast relying on MRI and as well as the advantages boosted by CT imaging have been combined by the using of MRI/CT. Unfortunately, disadvantages still exist in the MRI/CT workflow because the voxel-intensities are unbalanced in the MRI and the CT scan. Here, based on the mapping method of CT and MRI, the potential of pseudo-CT (PCT) instead of CT planning was studied. The estimated PCT only from the corresponding MRI was obtained by using the patch-based random forest regression. The CT voxel target was trained by 3D Gabor feature in the MRI cube and the Local Binary Pattern (LBP). Besides, the regression task was solved by the alternative regression forest. According to the experiment, the method performs better than the current dictionary learning-based (DLB) method or atlas-based (AB) method.

中文翻译：

---

仅 MRI 放射治疗：基于三次特征提取和替代回归森林的伪 CT

尽管磁共振成像 (MRI) 在放射治疗中引起了广泛关注，但研究人员重新引入了计算机断层扫描。在计算组织 3D 剂量分布的过程中，关于 CT 扫描获得的电子密度信息存在一些争论。然而，CT 提供的骨骼对于构建 X 光片来说是准确的。最近，依靠MRI的软组织对比增强的优势和CT成像增强的优势已经通过使用MRI/CT相结合。不幸的是，MRI/CT 工作流程中仍然存在缺点，因为 MRI 和 CT 扫描中的体素强度不平衡。在此，基于 CT 和 MRI 的映射方法，研究了伪 CT (PCT) 替代 CT 规划的潜力。通过使用基于补丁的随机森林回归获得仅来自相应 MRI 的估计 PCT。CT 体素目标通过 MRI 立方体中的 3D Gabor 特征和局部二值模式 (LBP) 进行训练。此外，回归任务由替代回归森林解决。根据实验，该方法的性能优于当前基于字典学习（DLB）的方法或基于图谱（AB）的方法。