【作者】 周绿漪； 苏显渝； 匡安仁；

【Author】 ZHOU Lu-yi ,SU Xian-yu,KUANG An-ren(Department of Nuclear Medicine, West China Hospital;School of Electronics and Information Engineering,Sichuan University, Chengdu of Sichuan Prov. 610041,China)

【机构】 四川大学华西医院核医学科； 四川大学电子信息学院；

【摘要】 研究γ照相机和SPECT像素尺寸测量中点源质心计算的优化方法。在点源图像上,以计数最大的像素为圆心生成一个包围点源的圆形区域。用区域内计数分布的质心代表点源质心,可在点源计效率较小的条件下,使本底计数分布对测量结果的影响减小。在一台状态良好的SPECT上应用本方法,并与原方法比较。在计数率范围为10-1183s-1时,本方法的测量结果为3.00-3.01mm(平均3.01±0.00mm),测量值与参考值的差异<1%。而原方法测量结果为3.10-23.11(平均7.07±6.35mm),测量值与参考值的差异很大。可见本方法的测量测结果不受点源活度的影响,而且有更高的准确度和精确度。更多还原

【Abstract】 Objective Pixel size is an important parameter of gamma camera and SPECT. A number of methods are used for its accurate measurement. In the original count-centroid method, where the image of a point source(PS) is acquired and its count-centroid calculated to represent PS position in the image, background counts are inevitable. Thus the measured count-centroid (X) is an approximation ofthe true count-centroid (Xp) of the PS, i.e. X = Xp + (Xb - Xp)/(1 + Rp/Rb), where Rp is the net counting rate of the PS, Xb the background count-centroid and Rb the background counting rate. To get accurate measurement, Rp must be very big, which is unpractical, resulting in the variation of measured pixel size. Rp-independent calculation of PS count-centroid is desired. Methods: The proposed method attempted to eliminate the effect of the term (Xb - Xp)/(1 + Rp/ Rb) by bringing Xb closer to Xp and by reducing Rb. In the acquired PS image, a circular ROI was generated to enclose the PS, the pixel with the maximum count being the center of the ROI. To choose the diameter (D) of the ROI, a Gaussian count distribution was assumed for the PS, accordingly, K=1-(0.5)D/R percent of the total PS counts was in the ROI, R being the full width at half maximum of the PS count distribution. D was set to be 6 * R to enclose most ( K = 98.4%) of the PS counts. The count-centroid of the ROI was calculated to represent X. The proposed method was tested in measuring the pixel size of a well-tuned SPECT, whose pixel size was estimated to be 3. 02mm according to its mechanical and electronic setting (128×128 matrix, 387mm UFOV, ZOOM = 1). For comparison, the original method, which was use in the former versions of some commercial SPECT software, was also tested. 12 PSs were prepared and their image acquired and stored. The net counting rate of the PSs increased from 10cps to 1183cps. Results: Using the proposed method, the measured pixel size (in mm) varied only between 3.00 and 3.01 (mean = 3.01±0.00) as Rp increased, and the difference between the measured and estimated pixel size was< 1% . While using the original method, the measured pixel size decreased from 23.11 to 3.10(mean = 7. 07±6.35) as Rp increased, and the difference between the measured and estimated pixel size was much bigger. Conclusions: The proposed method is not only Rp-independent, but also accurate and precise.更多还原