背景：氟－18-FDG (FDG)在腦內的分佈會隨時間而改變。本回溯性研究探討造影時間起始點之選擇對於影像分析的結果是否有影響。方法：32位健康受試者於注射FDG後40至60分鐘之間進行list mode掃描。我們選擇不同的起始時間，將影像資料重組爲兩組影像，兩組資料的擷取時間皆爲5分鐘，其一為注射FDG後第40至第45分鐘，另一組爲第55至第60分鐘。我們利用SPM 2 (statistical parametric mapping 2)統計軟體對這兩組腦部影像進行體素對體素(voxel to voxel)的放射活性分析比較，採用paired t-test找出兩組腦部PET影像之差異性。結果：若顯著性閥值設定在體素階層爲P值小於未校正0.001，在叢集階層爲P值小於已校正0.05，而相連體素所成之叢集中體素數量閥值設定爲300時，第40分鐘起始的腦部影像於右小腦前葉culmen、右小腦後葉pyramis、inferior semi-lunar lobule及tuber部份高於第55分鐘，其叢集中體素數量(K(下標 E))高達4,806個；第55分鐘起始的腦部影像於左腦額葉、middle frontal gyrus、limbic lobe、anterior cingulate，左腦顳葉、superior temporal gyrus，左腦頂葉、sub-gyral及右腦額葉sub-gyral、inferior frontal gyrus、middle frontal gyrus、precentral gyrus高於第40分鐘，其體素數量(E(下標 E))高達5,000個以上。然而當顯著性閥值設定較嚴格時，也就是體素階層設定爲P值小於已校正0.05，則兩組影像間的差異則較不顯著。結論：掃描開始時間點之選擇的確會影響影像分析之結果，但若統計之顯著性閥值設定較爲嚴格時，此種影響相對會變小。

Background: The distribution of 18F-FDG in brain changed with time. The purpose of this retrospective study is to find out the choice of emission acquisition starting-time may affect the resulting images. Methods: Thirty two healthy normal volunteers (mean age±SD, 54.8±10.9) received list mode 18F-FDG-PET scans starting from 40 to 60 min post radiotracer injection. We post-process the data into two 5-min images, one started from 40 to 45 minutes and the other started from 55 to 60 min post-injection. Voxel-to-voxel analysis using SPM 2 to compare the two sets of images. Results: If significant threshold of voxel-level set to 0.001 (uncorrected), cluster-level set to 0.05 (corrected) and cluster extent set to 300, 40 min images showed higher radioactivity than 55 min images within culmen, pyramis, inferior semi-lunar lobule and tuber of right cerebellum, the extent of cluster was 4,806 voxels. On the other hand, 55 min images showed higher radioactivity within bilateral frontal lobe, left medial frontal gyrus, left limbic lobe and left anterior cingulate gyrus, the extent of cluster was 5,132 voxels. If more severe threshold was set to voxel-level 0.05 (corrected), then very few differences was seen. Conclusion: The selection of starting time may influent the result of voxel-to-voxel analysis. If more severe threshold was set, the influence was much less.