The purpose of this study is to validate a previously developed algorithm for alerting clinicians when to consider re-CT simulation due to changes in the patient's anatomy during radiation therapy of head and neck cancer. Cone beam computed tomography (CBCT) data were collected prospectively for 77 patients. Each CBCT was mathematically compared to a reference CBCT using the gamma index. We defined the match quality parameter (MQP) as an indicator of CBCT image similarity, where a negative MQP value indicates a poorer CBCT match than the match between the first two CBCT acquired during treatment. If three consecutive MQP values were below a chosen threshold, an “alert” is triggered to indicate action required, for example, possible re-CT simulation. The timing of image review requests made by the radiation therapists and any re-CT/re-plan decisions were documented for each patient's treatment course. The MQP for each patient (including any re-plans) was calculated in a manner that was blinded from the clinical process. The MQP as a function of fraction number was compared to actual clinical decisions in the treatment progress to evaluate alert system performance. There was a total of 93 plans (including re-plans) with 34 positives (action required) and 59 negatives (no action required). The sensitivity of the alert system was 0.76 and the false positive rate was 0.37. Only 1 case out of the 34 positive cases would have been missed by both the alert system and our clinical process. Despite the false negatives and false positives, analysis of the timing of alert triggers showed that the alert system could have resulted in seven fewer clinical misses. The alert system has the potential to be a valuable tool to complement human judgment and to provide a quality assurance safeguard to help improve the delivery of radiation treatment of head and neck cancer.

中文翻译：

---

用于监测头颈癌放射治疗期间患者解剖结构变化的警报系统

本研究的目的是验证先前开发的算法，用于提醒临床医生何时考虑重新 CT 模拟，因为在头颈癌放射治疗期间患者的解剖结构发生了变化。前瞻性收集了 77 名患者的锥形束计算机断层扫描 (CBCT) 数据。每个 CBCT 都使用伽玛指数与参考 CBCT 进行数学比较。我们将匹配质量参数 (MQP) 定义为 CBCT 图像相似性的指标，其中负 MQP 值表示 CBCT 匹配比治疗期间获得的前两个 CBCT 之间的匹配差。如果三个连续的 MQP 值低于选定的阈值，则会触发“警报”以指示所需的操作，例如可能的重新 CT 模拟。放射治疗师提出的影像审查请求的时间安排以及任何重新 CT/重新计划的决定都记录在每个患者的治疗过程中。每个患者的 MQP（包括任何重新计划）的计算方式与临床过程不相关。将作为分数函数的 MQP 与治疗过程中的实际临床决策进行比较，以评估警报系统的性能。共有 93 个计划（包括重新计划），其中 34 个正面（需要采取行动）和 59 个负面（不需要采取行动）。警报系统的灵敏度为 0.76，误报率为 0.37。在 34 个阳性病例中，只有 1 个病例会被警报系统和我们的临床过程遗漏。尽管存在误报和误报，对警报触发时间的分析表明，警报系统本可以减少 7 次临床失误。警报系统有可能成为补充人类判断的宝贵工具，并提供质量保证保障，以帮助改善头颈癌放射治疗的实施。