摘要目的:探索递归分段-置换（RS-P）控制图在近距离放疗CT图像中对施源器金属伪影进行监测的可行性，并与单值移动极差（I-MR）控制图进行比较。方法:回顾性收集36例患者共181例图像用于过程分析。首先基于8种去金属伪影迭代算法（iMAR）预设参数及不使用iMAR算法重建其中15例得到135套图像，并由两名医师进行评分。随后提出了一种基于噪声功率谱（NPS）的NPS r指标以量化施源器相关的金属伪影严重程度（MAS），以评分为参考评估其性能。最后基于该指标分别采用RS-P控制图及I-MR控制图开展Phase I分析，识别、分析和剔除异常图像。 结果:两名医师的主观评分一致性良好（Kappa=0.86），所提出的NPS r指标与共识评分高度相关（ ρ s=0.92），可有效区分不同评分等级（ pf<0.01， pLMM<0.01），多个受试者特征曲线（ROC）的平均曲线下面积（AUC）为0.81。RS-P控制图有效检出了MAS的散在突变和由iMAR预设参数变化引起的系统性漂移，通过识别、剔除与重新计算获得了受控样本；而I-MR控制图对系统性变化不够敏感，且存在由异常值污染引发的误判。 结论:NPS r可有效量化施源器相关MAS，I-MR与RS-P控制图均适用于MAS的Phase I分析，但后者在系统性变化的检测上表现更优。

abstractsObjective:Exploring the Feasibility of Recursive Segmentation-Permutation (RS-P) Control Chart for Monitoring Applicator-Related Metal Artifacts in Brachytherapy CT Images and Comparing It with the Individual-Moving Range (I-MR) Control Chart.Methods:CT images from 36 patients (181 images) were retrospectively analysed. Fifteen images were additionally reconstructed with eight iterative metal-artefact-reduction (iMAR) presets and with no iMAR, yielding 135 data sets that were scored independently by two radiation oncologists. A noise-power-spectrum-based metric (NPS r) was proposed to quantify MAS severity and its performance was validated against the visual scores. Using NPS r as the quality characteristic, Phase-I process analysis was carried out with both RS-P and I-MR control charts to identify, investigate and remove out-of-control images. Results:The inter-rater agreement between the two physicians was good (Kappa=0.86). NPS r correlated strongly with the consensus scores ( ρs=0.92) and discriminated the score levels well ( pf<0.01; pLMM<0.01). The average area under the receiver operating characteristic curves (ROC AUC) across multiple binary classifications was 0.81. The RS-P chart detected both isolated MAS spikes and systematic step shifts caused by changes in iMAR presets, and iterative removal of assignable-cause images produced an in-control data set. In contrast, the I-MR chart was less sensitive to systematic shifts and showed false alarms due to outlier contamination. Conclusions:NPS r is an effective metric for quantifying source-related MAS. Although both I-MR and RS-P control charts are applicable to Phase-I analysis, RS-P demonstrates superior sensitivity and discriminatory power for detecting systematic changes.