Galvin J M, Sims C, Dominiak G, Cooper J S
New York University Medical Center, Division of Radiation Oncology, NY 10022.
Int J Radiat Oncol Biol Phys. 1995 Feb 15;31(4):935-42. doi: 10.1016/0360-3016(94)00503-6.
Recent interest in computed tomography-simulation (CT-simulation) suggests the possibility of a shift to digital images for field verification. This article examines the quality of Digitally Reconstructed Radiographs (DRRs) to determine if they can reasonably be substituted for conventional simulator films, and suggests techniques to improve these images.
Special developmental software and computer hardware allowed extremely rapid reformatting of CT data to produce images geometrically equivalent to treatment unit portal films. The technique uses a trilinear interpolation algorithm and gives a 512 x 512 DRR for any arbitrary beam direction. Resolution in line pairs/cm (lp/cm) for both simulator radiographs and DRRs was measured with a special test phantom. Patient data was reformatted to illustrate methods for improving the quality of the DRR.
The equipment used for this study reformats 50 512 x 512 CT scans in 8 s. The resolution for a DRR is limited by the voxel size of the CT scans. For typical voxel dimensions, the resolution was found to be 7 lp/cm transverse and 1.0 lp/cm longitudinal compared to 21.0 lp/cm for a simulator radiograph. Patient movement during the scan procedure further degrades the DRR. The reduced quality of this image makes it more difficult to discern structures, and it may not always be possible to perform essential tasks such as counting vertebral bodies. However, viewing the treatment field superimposed on a DRR displayed with a step function to include only bone aided in the identification of relevant landmarks. Switching between soft tissue, bone, or air windows takes less than 10 s on the equipment used for this study, and the use of different display techniques improved the viewer's ability to evaluate field positioning.
A DRR cannot match the spatial resolution of a radiograph taken with a short exposure and small focal spot, but the ability to change the display mode for the DRR increases the usefulness of these images. Fast reformatting is particularly important when evaluation of field position requires the comparison of each portal image to a series of DRRs.
近期对计算机断层扫描模拟（CT模拟）的关注表明，有可能转向使用数字图像进行射野验证。本文研究了数字重建射线照相（DRR）的质量，以确定它们是否能合理替代传统模拟定位机片，并提出改善这些图像的技术。
特殊的开发软件和计算机硬件允许对CT数据进行极其快速的重新格式化，以生成在几何上等同于治疗单元射野片的图像。该技术使用三线性插值算法，可针对任意射束方向生成512×512的DRR。使用特殊测试模体测量模拟定位机射线照片和DRR的线对/厘米（lp/cm）分辨率。对患者数据进行重新格式化，以说明提高DRR质量的方法。
本研究使用的设备在8秒内可对50幅512×512的CT扫描进行重新格式化。DRR的分辨率受CT扫描体素大小的限制。对于典型的体素尺寸，发现其横向分辨率为7 lp/cm，纵向分辨率为1.0 lp/cm，而模拟定位机射线照片的分辨率为21.0 lp/cm。扫描过程中患者的移动会进一步降低DRR的质量。这种图像质量的下降使得辨别结构更加困难，并且可能并非总能执行诸如计数椎体等基本任务。然而，将射野叠加在以阶梯函数显示的仅包含骨骼的DRR上有助于识别相关标记。在本研究使用的设备上，在软组织、骨骼或空气窗之间切换耗时不到10秒，并且使用不同的显示技术提高了观察者评估射野定位的能力。
DRR无法与短曝光和小焦点拍摄的射线照片的空间分辨率相匹配，但改变DRR显示模式的能力增加了这些图像的实用性。当评估射野位置需要将每个射野图像与一系列DRR进行比较时，快速重新格式化尤为重要。
