摘要目的:探讨肾脏囊性病变（CRM）2019版Bosniak分类中MRI标准的观察者重测信度，分析病灶病理性质、大小及阅片者经验对重测信度的影响。方法:回顾性连续收集2009年1月至2019年6月于解放军总医院第一医学中心行常规肾脏增强MRI并获得手术病理结果的207例CRM患者，共207个病灶。基于2019版Bosniak分类系统的MRI标准，8名不同年资的放射科医师在对患者病理信息均不知情的前提下，对所有CRM间隔1个月进行两次独立评估。采用组内相关系数（ICC）衡量8名阅片者对所有病灶以及不同病理性质（良性及恶性）、不同大小（最大径≤40 mm及>40 mm）亚组的重测信度，计算并比较4名高年资（工作经验≥10年）及4名低年资（工作经验<10年）阅片者的重测信度，ICC间比较采用Z检验。结果:207个病灶包括良性病灶111个（良性囊肿83个、良性肿瘤28个）和恶性肿瘤96个。最大径≤40 mm的87个，>40 mm的120个。基于2019版Bosniak分类MRI标准，对全部病灶，8名阅片者单独两次评估的重测信度ICC为0.776~0.888，整体重测信度ICC为0.848（95%CI 0.821~0.872），高、低年资阅片者重测信度ICC分别为0.853（95%CI 0.824~0.880）、0.843（95%CI 0.811~0.871），差异无统计学意义（ Z=0.85， P=0.374）。全部阅片者对良性病变的ICC为0.827，恶性病变为0.654，差异有统计学意义（ Z=2.80， P=0.005）；最大径≤40 mm的病变的ICC为0.770，>40 mm的为0.876，差异有统计学意义（ Z=-2.36， P=0.018）。对不同病理性质及不同大小的CRM亚组中，高、低年资阅片者重测信度差异均无统计学意义（ P均>0.05）。 结论:2019版Bosniak分类中MRI标准对CRM的阅片者重测信度佳，且不受阅片者工作经验影响；不同病理性质及不同大小病灶间重测信度存在差异，但均达良好以上水平。

abstractsObjective:To evaluate the test-retest reliability of MRI criteria in the 2019 Bosniak classification of cystic renal masses (CRMs) and to analyze the impact of lesions′ property, size and readers′ experience on the test-retest reliability.Methods:From January 2009 to June 2019, 207 patients with 207 CRMs were included in this retrospective study. All of them underwent renal MRI and surgical-pathologic examination. According to Bosniak classification, version 2019, all CRMs were independently classified twice by eight radiologists with different levels of experience. All radiologists were blinded to the pathology of the lesions. By using intraclass correlation coefficient (ICC), test-retest reliability was evaluated for all CRMs and for subgroups with different pathological properties (benign and malignant) and different sizes (≤40 mm and>40 mm). The test-retest reliability of 4 senior readers (≥10 years of experience) and 4 junior readers (<10 years of experience) were evaluated respectively. The comparison of ICC was performed using Z test. Results:The 207 CRMs included 111 benign lesions (83 benign cysts, 28 benign tumors) and 96 malignant tumors. There were 87 lesions with maximum diameter ≤40 mm and 120 with maximum diameter>40 mm. The test-retest reliability (ICC) of each reader for all lesions was 0.776-0.888, the overall ICC was 0.848 (95%CI 0.821-0.872). The ICCs of senior and junior readers were 0.853 (95%CI 0.824-0.880) and 0.843 (95%CI 0.811-0.871) respectively, without significant difference between the two groups ( Z=0.85, P=0.374). The ICC of all readers was 0.827 for benign lesions and 0.654 for malignant lesions, showing significant difference ( Z=2.80, P=0.005). The ICC was 0.770 for lesions ≤40 mm and 0.876 for lesions>40 mm, which was significantly different ( Z=-2.36, P=0.018). For CRM subgroups with different pathological properties and different sizes, there was no significant difference in test-retest reliability between senior and junior readers (all P>0.05). Conclusion:The test-retest reliability of MRI criteria in the 2019 Bosniak classification of CRMs is excellent and unaffected by readers′ experience. The reliabilities are not consistent among CRMs of different pathological properties and different sizes, but all reached the level of good and above.