Within-study repeated measurements to increase sensitivity for positron emission tomography activation studies

Abstract

A novel data collection strategy was examined for positron emission tomography activation studies. After an injection of H215O, data were collected in multiple 10-second frames and analyzed with a blocked analysis of variance design in which blocking was performed across frames. An estimate of residual error based on a larger number of statistically independent measurements was hence obtained and the statistical significance of detected differences increased. The feasibility of the suggested scheme was demonstrated on phantom data, where higher significance was achieved when dividing the same data into more frames. The method was further used for single-subject analysis of data from eight human subjects participating in a study on visceral sensation. The results show agreement with the group-based analysis and indicate that it is possible to detect areas with changes of 10 mL/(min · 100 mL) or more in single subjects. The residuals from the statistical analysis were analyzed and did not indicate any violations of the assumptions of statistical independence between frames, normal distribution of errors, and homoscedasticity across blocks. The specificity was worse than the theoretically expected 0.05, but this may have resulted from lack of complete control over the experimental situation rather than the statistical method per se.