^(18)F-FDG PET/CT SUL_(peak)测量方法可靠性分析

Reliability analysis of measurement methods for ^(18)F-FDG SUL_(peak) in PET/CT scan

目的分析手工及PET VCAR(PET Volume Computed Assisted Reading)软件自动测量肝脏及肺癌病灶SULpeak的重复性和一致性。方法回顾性分析50例原发性肺癌患者18 F-FDG PET/CT显像结果。两名医生、同一医生一周前后分别测定肝右后叶及肺癌病灶感兴趣区内高代谢区域的5个体积为1cm3(直径为1.2cm球形)的随机像素点的SULmax,取5个SULmax的均数定义为SULpeak-手工;同时应用AW46工作站的PET VCAR软件自动测定肝右后叶及肺癌病灶SULpeak-自动。采用Bland-Altman分析和配对T检验评价SULpeak-手工和SULpeak-自动的重复性和一致性。结果通过PET VCAR测量肝右后叶和肺癌病灶的SULpeak-自动误差均为0g/ml*,重复性均为100%。肝右后叶同一医生不同时间及不同医生手工测量的SULpeak-手工值均低于SULpeak-自动(P<0.001);Bland-Altman分析示同一医生不同时间及不同医生间分别有1/50(2%)、2/50(4%)个点位于95%一致性界限外。肺癌病灶SULpeak-手工值均高于SULpeak-自动(P<0.05);Bland-Altman分析示同一医生不同时间及不同医生间分别有1/50(2%)、2/50(4%)个点位于95%一致性界限外。结论SULpeak-手动受手工操作的影响明显,重复性和一致性均较SULpeak-自动差,而SULpeak-自动有良好的可靠性。PET VCAR自动测量SULpeak为PET/CT精确定量提供了重要平台,有助于推动以SULpeak为主要评价指标的PERCIST(Positron Emission tomography Response Criteria In Solid Tumors)标准的临床应用及推广。

Objective To compare peak standardized uptake normalized to lean body mass （SULpeak ） of manual and automatic measurement by PET VCAR （PET Volume Computed Assisted Reading ） and evaluate the repeatability of manual SULpeak. Methods Fifty patients with primary lung cancer confirmed by 18 FFDG PET/CTwere included in our study retrospectively. We used PET VCAR software of AW4. 6 workstation to automatically measure SULpeak （SULpeak -auto） and the average SULpeak of five random pixels in fixed sized （1.2 cm in diameter,about 1 cm3 volume） spherical VOI drawn over lesions was calculated by two doctors and same doctor at different time, defined as SULpeak-manual. Results The bias of SULpeak were 0 g/ml ＊ and repeatability was 100% for liver and lung cancer le- sions. While liver SULpeak-manual of two doctors and same doctor at different time were lower than SULpeak-auto （P〈0. 001）. And Bland-Altman analysis results showed that 1/50 （2 %） 2/50 （4 %） points located outside 95 % limit of agreement respectiveby. Lung cancer lesions SgLpoak-manual of two doctors and same doctor at different time were higher than SULpeak （P〈〈0.05）. And Bland-Airman analysis results showed that 1/50 （2 % ）, 2/50 （4 % ） points located outside 95 % limit of agreement respectivly. Conclusion The SULpeak-manual mainly depended on manual measurement,whose re peatability and consistence were inferior to SULpeak-auto, while the SULpeak-auto had good reliability. The accuracy and relia- bility of SULpeak-auto from PET VCAR could be suitable to assess treatment response, and be helpful to PERCIST clinical application and promotion,which evaluates response mainly by SULpeak