Background

Existing gated delivery methods adopting predefined intervention thresholds, and better delivery accuracy comes at the expense of lower therapeutic efficiency. So we present a new gating strategy based on trajectory quantification and pattern discrimination.Material and methods: 1265 real-time ultrasound monitoring trajectories of 61 prostate cancer patients were retrospectively analyzed. The trajectories included four typical patterns: stable target at baseline, transient excursion, continuous drift, and persistent excursion. Trajectories were randomly sampled to train a discriminant model that utilizes a sliding window approach with a 30-s window. The discriminant accuracy and receiver operating characteristic (ROC) curve were analyzed to assess the model’s performance.Results:The discriminant model to identify intra-fractional motion patterns was successfully trained. The model had statistical significance (p < 0.001) and correctly classified 94.0% of trajectories in the time window. The area under the ROC curve, sensitivity, and specificity were all greater than 0.80 for different motion patterns.Conclusion: As an essential component of the new gating strategy, trajectory quantification makes it possible to automatically discriminate intra-fractional motion patterns; pattern discrimination further allows active guidance for subsequent interventions. This preliminary work has shown promise in delivering radiotherapy accurately and cost-effectively.

中文翻译：

---

量化分数内前列腺运动轨迹以建立新的门控策略：初步研究

背景

现有的门控递送方法采用预定的干预阈值，并且更好的递送准确性是以较低的治疗效率为代价的。因此，我们提出了一种基于轨迹量化和模式识别的新门控策略。材料与方法：回顾性分析61例前列腺癌患者的1265次实时超声监测轨迹。轨迹包括四种典型模式：基线稳定目标，瞬时偏移，连续漂移和持续偏移。随机采样轨迹以训练判别模型，该模型使用带有30秒窗口的滑动窗口方法。分析判别精度和接收器工作特性（ROC）曲线以评估模型的性能。结果：识别分数内运动模式的判别模型已成功训练。该模型具有统计显着性（p <0.001），并在时间窗口中正确分类了94.0％的轨迹。对于不同的运动模式，ROC曲线下的面积，灵敏度和特异性均大于0.80。结论：作为新选通策略的重要组成部分，轨迹量化使自动区分分数内运动模式成为可能。模式识别还可以为后续干预提供积极的指导。这项初步工作已显示出有望准确，经济地进行放射治疗。