The long-term objectives of this proposal are to develop image reconstruction and data analysis techniques for increasing the biological relevance of quantitative positron emission tomographic (PET) images of the human brain. The goal is to optimize the detection and measurement of PET information, with specific reference to the activation of functional brain networks and related brain/behavior relationships in normal and pathological states. Such optimized analysis techniques may provide sensitive diagnostic tools for quantifying disease progression and treatment. It has been shown as MSKCC that information on functional brain networks (1) is contained in a small fraction, is less than 20%, of absolute measurements of regional cerebral metabolic rates for glucose (Tau CMRG1u) and (2) may be detected and measured using a scaled subprofile model with factor analysis of variance (SSM/FANOVA). Using computer simulations of PET image data, PET images from phantom experiments, groups of normal and diseased Tau CRMG1u images and images of normal tissue blood flow, the aims of this proposal are to (1) characterize the statistical power and robustness of SSM/FANOVA compared to standard statistical analysis techniques (e.g., t-test, analysis of variance) for the detection and measurement of the functionally important Tau CMRG1u component; (2) optimize the detection and measurement of functionally related covariance patterns in Tau CMRG1u images by varying analysis variables, (e.g., the number of scans in a group and region-of-interest specification techniques), and possible trade-offs among the corrections for the wide range of quantitative PET error sources if absolute image singal accuracy is no longer a primary requirement; (3) test the hypothesis that increased PET scanner resolution will confound the analysis of small functional signal components in groups of PET images; (4) test hypothesis that the test-retest scanning protocol possible with tissue blood flow scans has statistical advantages for the measurement of a functionally related signal component and (5) develop new graphical and image display techniques based on an SSM/FANOVA analysis of small signal covariance patterns.
