CAJ | 학술논문

目的通过模型实验初步建立使18F-脱氧葡萄糖(FDG)PET图像上显示的肿瘤体积肿瘤实际体积一致的最佳阈值条件.方法采用美国DATA SPECTUM公司NEMA IEC-BodyPhantom setTM>体模,以不同大小的球体(体积分别为0.52,1.15,2.60,5.50,11.40,26.50 ml)模拟临床常见肺肿瘤的大小和靶/本底(T/B)比值.选择合适的窗位和窗宽,使CT图像上显示的球体体积与球体实测体积一致,并以此作为球体的实际体积,调节PET图像的阈值,使PET图像上勾画的球体轮廓与CT图像上所勾画的最接近.设绝对阈值(Ithreshold)=本底平均像素值(Iback)+相对阈值[x%Imax-back(slice)>,即每个层面的最大像素值与本底平均像素值的差值的百分比].比较应用筛选出的最佳绝对阈值与应用最广泛的靶体内最大像素值[42%Imax(total)>]为阈值勾画的球体体积大小及对实际体积覆盖率的差异.采用SPSS 13.0软件进行统计学处理,行Wilcoxon法检验.结果对体积≥5.50 ml的不同T/B比值的球体,中央1/2层面的最佳相对阈值与最大层面的最佳相对阈值相差(1.6±2.0)%,而两端1/2层面的最佳相对阈值与最大层面的最佳相对阈值相差(19.3±10.1)%,二者差异有统计学意义(Z=-7.26,P<0.01).最大层面的最佳相对阈值在20%～25%之间的改变引起勾画体积与实际体积的差异≤5%.以Ithreshold=Iback+20%Imax-back作为各层面的最佳绝对阈值,所勾画的球体体积较实际体积扩大(12.6±6.6)%,而42%Imax(total)勾画的体积较实际体积缩小(-17.2±15.2)%(Z=2.67,P<0.01),前者对实际体积的覆盖率优于后者(97.1%与79.4%,Z=-2.67,P<0.01).结论 Iback+20%Imax-back(slice)可能是不依赖于CT提供的肿瘤体积信息层面化的最佳阈值条件,其准确性优于42%Imax(total)阈值法.
목적 통과모형실험초보건립사18F-탈양포도당(FDG)PET도상상현시적종류체적종류실제체적일치적최가역치조건.방법 채용미국DATA SPECTUM공사NEMA IEC-BodyPhantom setTM>체모,이불동대소적구체(체적분별위0.52,1.15,2.60,5.50,11.40,26.50 ml)모의림상상견폐종류적대소화파/본저(T/B)비치.선택합괄적창위화창관,사CT도상상현시적구체체적여구체실측체적일치,병이차작위구체적실제체적,조절PET도상적역치,사PET도상상구화적구체륜곽여CT도상상소구화적최접근.설절대역치(Ithreshold)=본저평균상소치(Iback)+상대역치[x%Imax-back(slice)>,즉매개층면적최대상소치여본저평균상소치적차치적백분비].비교응용사선출적최가절대역치여응용최엄범적파체내최대상소치[42%Imax(total)>]위역치구화적구체체적대소급대실제체적복개솔적차이.채용SPSS 13.0연건진행통계학처리,행Wilcoxon법검험.결과 대체적≥5.50 ml적불동T/B비치적구체,중앙1/2층면적최가상대역치여최대층면적최가상대역치상차(1.6±2.0)%,이량단1/2층면적최가상대역치여최대층면적최가상대역치상차(19.3±10.1)%,이자차이유통계학의의(Z=-7.26,P<0.01).최대층면적최가상대역치재20%～25%지간적개변인기구화체적여실제체적적차이≤5%.이Ithreshold=Iback+20%Imax-back작위각층면적최가절대역치,소구화적구체체적교실제체적확대(12.6±6.6)%,이42%Imax(total)구화적체적교실제체적축소(-17.2±15.2)%(Z=2.67,P<0.01),전자대실제체적적복개솔우우후자(97.1%여79.4%,Z=-2.67,P<0.01).결론 Iback+20%Imax-back(slice)가능시불의뢰우CT제공적종류체적신식층면화적최가역치조건,기준학성우우42%Imax(total)역치법.
Objective In the literature, data regarding the optimal threshold for defining the true target volume in radiotherapy are controversial. The aim of this study was to use a phantom to evaluate an optimal threshold segmentation for 18F-fluorodeoxyglucose (FDG) PET/CT clinical applications. Methods A NEMA IEC-Body SetTM phantom (DATA SPECTUM, USA) with spheres in various sizes (volume: 0.52, 1.15, 2.60, 5.50, 11.40 and 26.50 ml) was used to simulate lung cancer lesions of different tar-get/background (T/B) ratios. Firstly, an appropriate window in CT slice was selected and adjusted so that the sphere volumes on CT are equal to the true sphere volumes on phantoms. Secondly, the threshold of PET images was adjusted to an optimal value such that the PET volume equalizes the CT volume. An optimal threshold was thus calculated by the following formula: Ithreshold=Iback>+x% Imax-back(slice), where Ithreshold Was an absolute threshold, Iback was mean radioactivity in background, and x% Imax-back(slice) was a relative threshold. Both the size and coverage delineated at optimal threshold were compared with those generated at a fixed threshold of 42% maximum voxel radioactivity (commonly used in commercial programs). SPSS 13.0 and Wilcoxon test were used for statistical analysis. Results For spheres > 5.50 ml with various T/B ratios,the mean deviation of optima] threshold between the middle half slices and the maximum slice was (1.6±2.0)%, whereas the corresponding value between the peripheral half slices and the maximum was (19.3±10.1)% (Z=-7.26, P <0.01). The % volume-deviation calculated with different optimal thresholds at maximum slice was ≤5% if the relative threshold varied within 20%-25%. The mean % vulume-deviation delineated by Ithreshold= Iback+20% Imax-back(slice) Was (12.6±6.6)%, while the corresponding value delinea-ted by a fixed threshold of 42% Imax(total) Was (-17.2±15.2) % (Z=-2.67, P<0.01). These results implied that Ithreshold = Iback+20% Imax-back(slice)> was superior to 42% Imax(total) in the estimation of spherical le-sion volume (97.1% vs 79.4%, Z=-2.67, P<0.01). Conclusion Iback+20% Imax-back(slice)> might be a more optimal threshold segmentation method than the conventional 42% Imax(total) parameter for more accurate target volume delineation and is independent of the target volume information provided by CT slice in ad-vance.