AI Denoising for Enhancing Gated Parametric Mapping of Rubidium-82 Cardiac PET

Cardiac PET is a powerful tool to assess heart function. One measure from a Cardiac PET study is left ventricle ejection fraction. This is mainly estimated through gated reconstructions of relative perfusion imaging and segmentation of the left ventricle myocardium. However, direct estimations of left ventricle volume and thus ejection fraction can be obtained by performing pharmacokinetic modeling on gated dynamic reconstructions to yield gated parametric images of blood fraction. However, gated dynamic reconstructions are subject to more noise due to binning short time frames into 8 or 16 bins throughout the cardiac cycle. The intern in this project will apply AI tools to denoise 1/8 or 1/16th dose images to full dose images (on short time frames) in order to improve pharmacokinetic fits and thus higher quality images of gated blood fraction and more accurate measures of ejection fraction.

Faculty Supervisor:

Robert DeKemp

Student:

Partner:

Jubilant Radiopharma

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

University of Ottawa

Program: