评估读片者经验对多参数前列腺MRI中PI-RADS 2.1版3分评分一致性的影响:单中心分析
本单中心回顾性研究纳入133例PI-RADS 3分且有病理证实的患者,由3名不同经验水平的放射科医生独立评分。读片者间一致性总体较差(中位κ=0.149),PI-RADS与Gleason评分仅弱正相关且无统计学显著性(τ=0.197),不同经验水平未显著影响PI-RADS 3评估。
要点
- PI-RADS 2.1版3分病灶的读片者间一致性总体较差(中位κ=0.149)。
- 不同经验水平放射科医生对PI-RADS 3的评估未见显著差异。
- Senior Reader 2的AUC(0.748)、准确率(63.2%)、PPV(48.2%)、NPV(88.0%)和敏感度(86.97%)最高。
核心结论
- 三名不同经验水平的放射科医生使用PI-RADS 2.1版指南的读片者间一致性总体较差,不同专业经验未显著影响PI-RADS 3评估;作者提出未来应改进判读流程、辅助决策规则、检验加权一致性统计量并在更大规模多中心队列中验证这些发现。
研究设计与样本
- 单中心回顾性研究,2023年1月至2025年4月共673例患者,排除既往干预、非PI-RADS 3分或缺乏病理证实的病例后纳入133例分析;由3名不同经验水平的放射科医生独立评估多参数前列腺MRI图像,以病理结果为参考标准。
诊断标准与关键数值
- 读片者间一致性采用Cohen kappa统计量评估,所有读片者两两之间的中位κ为0.149;PI-RADS与Gleason评分相关性采用Kendall tau分析,τ=0.197,呈弱正相关但无统计学显著性;Senior Reader 2的AUC为0.748、准确率63.2%、PPV 48.2%、NPV 88.0%、敏感度86.97%;Expert Reader 1的特异度最高,为56.3%。
临床应用与阅片要点
- 作者提出未来应改进判读流程、辅助决策规则并检验加权一致性统计量。
- 作者建议在更大规模多中心队列中验证这些发现。
局限性
- 原文明确为单中心分析。
- 作者指出未来需在更大规模多中心队列中验证这些发现。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
PURPOSE: PI-RADS 3 lesions remain a "grey-zone" with an equivocal category on prostate mpMRI, complicating biopsy decisions and treatment strategies. This study aimed to investigate how radiologist experience affects the interpretation of prostate multiparametric magnetic resonance imaging (mpMRI) with Prostate Imaging Reporting and Data System (PI-RADS) score 3 of version 2.1.
METHODS: A retrospective study involved three radiologists of different experience levels independently evaluating mpMRI images with PI-RADS 3 scores and confirmed pathological outcomes. From January 2023 to April 2025, 673 patient cases were reviewed. Excluding cases with prior interventions, non-PI-RADS 3 scores, or lacking pathological confirmation, 133 patients were analyzed. Inter-reader reliability was assessed using Cohen's kappa statistic (k), and correlations between PI-RADS and Gleason scores were analyzed with Kendall's tau (t). Diagnostic performance metrics including the area under the ROC curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated.
RESULTS: Median k score among all radiologist pairs was 0.149. The correlation between PI-RADS and Gleason scores was weakly positive but not statistically significant (t = 0.197). Senior Reader 2 demonstrated the highest AUC (0.748), accuracy (63.2%), PPV (48.2%), NPV (88.0%), and sensitivity (86.97%). Expert Reader 1 had the highest specificity (56.3%).
CONCLUSION: Three assigned radiologists with different experience levels achieved generally poor inter‑reader agreement using the PI‑RADS version 2.1 guideline, and varying expertise did not significantly influence PI-RADS 3 evaluations. Future work should refine interpretation flows, adjunct decision rules, test weighted agreement statistics, and validate these findings in larger, multicenter cohorts.
引用
- He M, Rzayev R, Serova N, et al. Evaluating the impact of reader experience on PI-RADS 3 of version 2.1 scoring concordance in multiparametric prostate MRI: a single-center analysis. Abdom Radiol (NY). 2025. PMID: 40996515. doi:10.1007/s00261-025-05207-0
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。