Longitudinal study of changes in FDDNP-PET and neuropsychological scores

Objectives To predict changes in neuropsychological test scores from longitudinal distribution volume ratio (DVR) images of FDDNP-PET data.

Methods 14 subjects (6 Control, 8 MCI) underwent a dynamic FDDNP scan for up to 125 minutes. T1-weighted MPRage MRIs were performed for each subject. FDDNP scans were repeated for all subjects two years later. At both visits, subjects were given a battery of neuropsychological exams. These tests included MMSE, Boston Naming, WAIS-III Block Design, and WAIS-III Digit Symbol tests. Logan analysis was performed to create DVR images for all subjects for both FDDNP scans. Subject MRIs were normalized to a common space through of structural warping. This normalization was applied to rigidly aligned DVR images to bring them into the common space. VOIs drawn in the common space were used to extract average DVR values from 9 cortical regions in the normalized PET images. Multiple linear regression was used to model changes in psychological test scores as a function of changes in regional DVR values. An exhaustive search of models using three or fewer predictor regions was performed to find the best models.

Results Block Design scores best correlated with data from frontal lobe, lower temporal, and posterior cingulate regions (r²=.87, p<0.001) that are related to planning and execution, memory, and integration, respectively. Boston naming test scores were related to upper parietal, lower temporal, and medial temporal regions (r²=.65, p=0.011), corresponding to the use of association cortices and memory. Digit symbol scores correlated with data in occipital-parietal, posterior cingulate, and lower temporal regions (r²=.67, p=0.015), which are used in association, integration, and memory. MMSE correlated with data from upper parietal and posterior temporal regions (r²=.80, p<0.001). In all cases, increase in DVR was inversely correlated with test performance.

Conclusions Changes in regional FDDNP-DVR data can be used to predict changes in specific mental tasks associated with the corresponding regions.

Research Support NIH grant P01-AG025483