Automated brain parcellation and morphometry

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Alzheimer's disease (AD) is a devastating disease that is reaching epidemic proportions as the US population ages. Early diagnosis of AD is crucial for individuals and families to plan for their futures, to obtain appropriate care, and to start treatment as early as possible. Despite this clear and pressing need, early AD is severely under-diagnosed. Research studies show that quantification of cerebral atrophy from MRI could aid in AD diagnosis; the current grant proposes to complete the steps necessary for FDA regulatory approval of a software device that could provide such information automatically to clinicians. Our first aim is to integrate, refine and further optimize the structural analysis stream developed in previous SBIR grants. This will involve porting research code into an FDA-approvable code base, refinement of skull- stripping and partial volume estimation methods, and their integration into the segmentation processing stream software. We will implement a method for identifying cases where the automatic quantification may have yielded inaccurate results, and refine discriminant functions to optimally classify different diagnostic categories. Our second aim is to modify our existing image pipeline software for clinical utilization as a robust commercial .product, adding industry standard (DICOM) services, inbound sequence parameter filtering and error/exception/management reporting, and an effective graphical interface for rapid clinical review, all within a completely automated processing flow integrated in a radiology department's digital environment. Our third aim is to evaluate the safety and efficacy this medical software device in order to obtain FDA Pre- Market Approval for its clinical use as an adjunct in the diagnosis of AD. This will require compliance with all relevant FDA regulations and guidelines, blinded laboratory testing and scientific validation of the automated quantitative segmentation results, clinical validation of user labeling (instructions and user interface usability) through external beta testing, and statistical validation of diagnostic labeling by using the stream in a well- controlled clinical population with AD. If successful, the current project would result in an FDA-approved product that will be widely used to increase the accurate and early diagnosis of AD.