This project aims to implement a database of neuropsychiatric PET images along with user-friendly, web-integrated tools to support neuropsychiatric clinical diagnosis and knowledge discovery (KDD). In Phase I, we seek to provide solutions to three mediate and compelling needs. First, to make fully informed diagnoses and prognoses, clinicians now require fast and easy access to normative databases to interpret, beyond routine visual inspection, functional brain images of individual patients. Second, sensitive and efficient searches, that have the power to control for demographic variables, call for large datasets of both normal subjects and patients. Neuroimaging databases do not typically contain subjects screened in detail for neuropsychiatric disease. Such screening is essential given the 32% lifetime prevalence of mental illnesses -- resulting in long- term structural and functional cnsequences. Third, improved and more powerful database management systems (DBMS) and specialized computer interface must support such datasets and their query. We propose the following Specific Aims: 1) To implement a relational, interoperable, web-distributed database containing PET image data, procedures, and results, along with general tools for visualization and analysis. 2) To incorporate into the database our extant data (approximately 6,000 PET scans) for archiving, storage, curation, and KDD; and to acquire (for clinical application) fifty normative glucose metabolic scans (FDG). 3) To develop a clinician?s interface; and 4) To evaluate Phase I results a) by specific application to the diagnosis of Alzheimer?s disease (AD); by testing the generalizability of reference datasets across facilities; and by crossvalidating with respect to clinical follow-up and autopsy. This project should provide the necessary infrastructure and validation to proceed in Phase II toward a web-distributed federation of databases containing control and neuropsychiatric data to support clinical care and research.
