Institute of Information and Communication Technologies, Université catholique de Louvain, Louvain-La-Neuve, 1348, Belgium.
Proton Therapy Unit, Hospital of Trento, Trento, 38122, Italy.
Med Phys. 2017 Oct;44(10):5393-5401. doi: 10.1002/mp.12497. Epub 2017 Aug 31.
Proton radiography seems to be a promising tool for assessing the quality of the stopping power computation in proton therapy. However, range error maps obtained on the basis of proton radiographs are very sensitive to small misalignment between the planning CT and the proton radiography acquisitions. In order to be able to mitigate misalignment in postprocessing, the authors implemented a fast method for registration between pencil proton radiography data obtained with a multilayer ionization chamber (MLIC) and an X-ray CT acquired on a head phantom.
The registration was performed by optimizing a cost function which performs a comparison between the acquired data and simulated integral depth-dose curves. Two methodologies were considered, one based on dual orthogonal projections and the other one on a single projection. For each methodology, the robustness of the registration algorithm with respect to three confounding factors (measurement noise, CT calibration errors, and spot spacing) was investigated by testing the accuracy of the method through simulations based on a CT scan of a head phantom.
The present registration method showed robust convergence towards the optimal solution. For the level of measurement noise and the uncertainty in the stopping power computation expected in proton radiography using a MLIC, the accuracy appeared to be better than 0.3° for angles and 0.3 mm for translations by use of the appropriate cost function. The spot spacing analysis showed that a spacing larger than the 5 mm used by other authors for the investigation of a MLIC for proton radiography led to results with absolute accuracy better than 0.3° for angles and 1 mm for translations when orthogonal proton radiographs were fed into the algorithm. In the case of a single projection, 6 mm was the largest spot spacing presenting an acceptable registration accuracy.
For registration of proton radiography data with X-ray CT, the use of a direct ray-tracing algorithm to compute sums of squared differences and corrections of range errors showed very good accuracy and robustness with respect to three confounding factors: measurement noise, calibration error, and spot spacing. It is therefore a suitable algorithm to use in the in vivo range verification framework, allowing to separate in postprocessing the proton range uncertainty due to setup errors from the other sources of uncertainty.
质子射线照相术似乎是评估质子治疗中停止能力计算质量的一种很有前途的工具。然而，基于质子射线照相术获得的射程误差图对计划 CT 与质子射线照相术采集之间的微小不对准非常敏感。为了能够在后期处理中减轻不对准的影响，作者实现了一种快速方法，用于在使用多层电离室（MLIC）获得的铅笔质子射线照相数据与头部体模上获得的 X 射线 CT 之间进行配准。
配准通过优化成本函数来执行，该函数对获得的数据与模拟积分深度剂量曲线进行比较。考虑了两种方法，一种基于双正交投影，另一种基于单个投影。对于每种方法，通过基于头部体模 CT 扫描的模拟测试该注册算法对三个混杂因素（测量噪声、CT 校准误差和斑点间距）的鲁棒性，来评估方法的准确性。
本注册方法显示出对最佳解决方案的稳健收敛性。对于在使用 MLIC 的质子射线照相术中预期的测量噪声水平和停止能力计算的不确定性，在使用适当的成本函数时，精度似乎优于 0.3°的角度和 0.3 毫米的平移。斑点间距分析表明，当将正交质子射线照相术输入到算法中时，对于大于其他作者用于研究 MLIC 用于质子射线照相术的 5 毫米的斑点间距，角度的绝对精度优于 0.3°，平移精度优于 1 毫米。在单个投影的情况下，最大斑点间距为 6 毫米，可提供可接受的注册精度。
对于 X 射线 CT 与质子射线照相术数据的配准，使用直接射线追踪算法来计算平方和差异的和并校正射程误差，对于三个混杂因素（测量噪声、校准误差和斑点间距）具有非常好的准确性和鲁棒性。因此，它是一种适合在体内射程验证框架中使用的算法，允许在后处理中分离由于设置误差引起的质子射程不确定性与其他不确定性源。
