CAREER: Developing a list-mode imaging paradigm

Washington University, Saint Louis MO

Investigators

Abstract

Imaging sciences have revolutionized discoveries in a multitude of scientific applications. Multiple imaging systems, which include systems for imaging living systems and those for gamma-ray astronomy, have the unique ability to acquire data on a per-photon basis and store this data in list format. However, the full potential of these systems, referred to as list-mode systems, remains untapped since current methods to process data from these systems are sub-optimal and result in information loss. This information loss is even more concerning for list-mode systems, since typically these systems are photon-starved, and it is vital that maximal information be extracted from each detected photon. Towards addressing this important need, in this CAREER project, a new paradigm to process data from list-mode systems will be developed. The development of this paradigm will open new frontiers on imaging small-sized regions and low-count imaging. Direct impact will be demonstrated in quantifying physiological properties of small structures deep in the brain towards the goal of understanding the pathophysiology of Parkinson disease. The paradigm is also poised to impact many other sciences where such systems are used including astrophysics, biomedical research, geoscientific process monitoring, and nuclear security and safety. The highly integrated educational objective will be to train, educate, and motivate students at all levels about the fundamental aspects of imaging science, with the goal of grooming the next generation of imaging scientists who possess strong mathematical proficiency and the ability to translate this proficiency to develop methods to process data from imaging systems. The proposed program aims to push the fundamental limits of imaging by developing novel theoretical and computational methods to extract task-specific information from list-mode imaging systems. Conventional computational imaging methods are typically designed to operate with discrete data. Thus, when processing data from list-mode systems, these methods first bin that data, which, as has been shown in multiple studies, results in loss of task-specific information. To avoid this information loss, there is a crucial need for new methods that can process the list-mode data in this continuous format. The intellectual significance of this proposal stems from this need and lies in the development of a continuous-to-continuous operator-based paradigm and associated methods to process list-mode data. The research approach is to (a) develop new information-theory-based techniques to quantify limits on task performance with list-mode systems, (b) design and validate algorithms to extract task-specific information from list-mode systems, including new reconstruction algorithms that will estimate continuous representation of the underlying object and (c) develop algorithms to estimate mean regional uptake in small regions of the brain from single-photon emission computed tomography images. The proposed list-mode paradigm will strongly impact both the fundamental and applied aspects of imaging science by providing new rigorous mathematical formalisms to process list-mode data, thereby yielding the ability to retrieve previously unextracted information from these systems. 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.