PET图像定量计算中的误差分析

Analysis of Noise on PET Images

目的 :分析PET影像中的误差成分 ,以期为分析PET定量计算精度提供思路。材料与方法 :PET ,质控模型 ,1 8F FDG ,在 5个剂量用 2D方式采集数据。计算1 8F FDG均匀分布区域像素计数均值和标准差 ,曲线拟合。结果 :综合误差平方随信号变化呈二次曲线型。误差成分可分为三部分 :( 1)与随机符合相关的成分 :正比于信号强度 ;( 2 )与放射性统计涨落相关的成分 :正比于信号强度平方根 ;( 3)其它成分 :与信号强度无关。结论 :和γ照相机和SPECT图像中的信噪比随信号源强度一直增加不同 。

Purpose: In order to properly conduct quantitative measurement with PET, components of noise were studied. Materials and Methods: ECAT EXACT HR +PET (Siemens) scanner and a quality control phantom (Toshiba) was used. Image data was collected in 2D mode at five doses of 18 F FDG. Pixel mean counts and standard deviation were calculated within ROI in the uniformity area of the phantom. Curve fitting was done with the data. Results: The square of synthetic noise on a PET image varies as signal and the relationship could be expressed with quadrative polynomial;  2=0 0395 N 2+0 1427 N+0 0025 (R 2=0 9358). The synthetic noise may be categorized into three components. The first one varies directly with signal intensity and is related to random coincidence; the second one was related to the square root of signal intensity and was related to Poisson fluctuation; and the third one does not relate to signal intensity. Conclusion: The noise arise mainly from the second component related to Poisson fluctuation; but with the increase of signal intensity, the proportion of the second component decreases and the proportion of the first component related to random conicidence increases.