This SBIR Phase I Project will productize a quantitative data processing technology in dynamic positron emission tomography (PET). Dynamic PET's distinction from its static counterpart is in the fact that it collects time-dependent tracer concentration data. It is a highly sensitive and accurate way of functional and molecular imaging of protein targets such as biomarkers of Alzheimer's disease, Parkinson's disease, or brain trauma. Unfortunately, dynamic PET is rarely used in clinic because of long imaging time and complex processing. The technology to be developed in this proposal will shorten the required imaging time and simplify the analysis, allowing for accurate and reliable imaging of disease biomarkers for aiding drug development, clinical diagnostics, and treatment monitoring. The proposed analytic workflow will be offered as a cloud-based service that can be used by the customers to analyze new or historical PET datasets. In order to overcome the established practice of relying on less accurate and informative static PET imaging, the grant recipients plan to conduct a thorough validation of the new technology using a dedicated physical phantom and human subject images acquired in the past. The largest impact of the new technology is expected in medical research and drug development.
The accepted practice of analyzing dynamic PET datasets involves reconstruction of the parameters of implied compartment model describing the tracer biochemistry, which involves several accuracy-degrading approximations and requires lengthy acquisition. The alternative approach proposed in this project relies on decomposing the dynamic PET image sequences using a combination of several processing techniques that incorporate the essential features of tracer pharmacokinetics without making explicit assumptions about the model parameters or about the underlying anatomy. The key part of the proposed methodology is factor analysis of dynamic structures, a non-negative matrix decomposition technique with successful past applications in radionuclide image analysis. The result of the workflow is a 3D concentration of tissues that exhibit specific-binding of the radiotracer. In order to validate the quantitative accuracy of the approach, the researchers will build and image a dedicated physical phantom capable of producing overlapping distributions with different tracer dynamics. In order to validate the scan-time reduction claims, the new workflow will be used to analyze existing dynamic PET human subject datasets using increasingly shorter input time windows.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
