Abstract: Objective To investigate the value of statistical parametric mapping 5 (SPM5) and its parameter settings in analysis of PET imaging for epilepsy patients. Methods Seventeen epilepsy patients and seventeen controls were scanned with PET. The datas were analyzed using PET and SPM5 with qualitative and semiquantitative analysis, statistical analysis of the percentage (numbers of agent distribution decreasing in focus to brain) for different parameters combinations, which were value adjustment to contral-p and voxels-k. Results There was decreasing of agent distribution at temporal lobe in epilepsy patients detecting with PET; the T/NT of focus to cerebellum was 1.07±0.24, the T/NT of normal tissue in opposite side to cerebellum was 1.27±0.18, there was a significant difference (t=1.87, P<0.05). Dealing with SPM5, there was more significant decreasing of agent in regions 19, 21, 39 of temporal lobe, regions 7, 19, 31, 40, 47 of parietal lobe and occipital lobe in epilepsy patients. With different value of k, the percentage was (27 ±22)%, (29±24)%, (32±23)%, (35±27)%, (39 ±31)% respectively, there was not significant difference (F=0.59, P>0.05); for different value of p, the percentage was (42±30)%, (29±25)%, (26±21)% respectively, with a significant difference (F=3.60, P<0.05); there was linear regression in value adjustment to contral-p and the percentage(b=-18.24, t=2.57, P<0.05). Conclusions Semiquantitative analysis by SPM5, the setting of value adjustment to contral-p would affect the results, the smaller of value adjustment to contral-p, the better of the result. SPM5 would be more objective and accurate to locate the focuses.