Statistical Error Analysis Based on NonParametric Bootstrap Method of Quantitative MBF Measurement with H/sub 2//sup 15/O_PET

Knowledg of the precision in the physiological parameters estimated by positron emission tomography (PET) is helpful for optimizing PET study and accurate diagnosis. Nonparametric bootstrap method proposed by Buvat is re- sampling techniques that can be used to accurately estimate the statistical properties of PET images in one scan and determine statistics of pixel values. The standard deviation (SD) of time activity curves (TAC) generated by region of interests (ROI) also are estimated by using the bootstrap method and the SD would propagate to estimated parameters such as blood flow. In order to evaluate the present method, a PET study for myocardial blood flow (MBF) measurement with H2 15 O_PET was performed. The statistical properties of MBF were estimated and the effects of scan-duration time and ROI size on the SD of MBF were developed. statistical properties of PET images in a single scan. The bootstrap approach is a computer-based statistical method for determining the accuracy of a statistic from experimental data (6). Buvat estimated statistical properties for PET and single photon emission computed tomography (SPECT) successfully using non-parametric bootstrap method (7). Noise estimates in reconstructed images could be introduced into quantitative data analysis to account for the uncertainty of pixel and region of interest (ROI) values. This study investigated the precision in the MBF with H2 15 O_PET using statistical properties expressed by bootstrapped images of a single acquisition. Using the concept of the non-parametric bootstrap method, we developed a method to know the standard deviation (SD) and coefficient of variation (COV) of estimated parameters of MBF, PTF, and arterial blood volume with H2 15 O_PET. A PET study with one pig was performed in order to evaluate the present technique. Strategies for scan duration and ROI selection were also investigated with the effect of SD and COV indicated by present method.