Radiation exposure of patients during medical imaging has become a major concern, with computed tomography (CT) and cardiac single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) being the biggest contributors. In this project our aim will be to dramatically reduce radiation dose in cardiac SPECT MPI based on inexpensive software techniques that can be translated readily to clinical practice. Our initial results suggest that at least an eightfold reduction in radiation doe might be possible, which could lead to an estimated reduction of 6500 cancer deaths per year in the U.S. alone. In lay terms, cardiac SPECT MPI is used routinely to evaluate heart disease, allowing the physician to assess blood flow reaching the heart wall, the motion of the heart, and whether the heart wall's tissue is viable. This form of imaging involves administration of a radioactive pharmaceutical (tracer) to the patient, which exposes the patient to radiation; thus, i would be desirable to minimize tracer dose used during imaging. Image quality is determined by the amount of tracer used, and these levels were chosen prior to recent technological improvements that can produce high-quality images at lower dose; therefore, there is clear evidence that doses can be lowered, and indeed there is considerable current interest in doing so. We are proposing a translational research project to thoroughly and quantitatively study the extent to which administered dose might be reduced without sacrificing image quality. This work will build, in particular, on new four-dimensional (4D) image reconstruction techniques and respiratory motion compensation approaches we have developed. Furthermore, we propose a new dosing approach we call personalized imaging, in which a computer algorithm will be trained to predict, for a given patient, the minimum dose required to obtain the current level of image quality, so that the administered dose is no more than necessary. In 4D reconstruction, a computer algorithm tracks the heart's beating motion, and uses this information to improve image quality by smoothing image noise in a way that preserves image details. In respiratory motion compensation, the patient's breathing and body motions are tracked by external sensors, and this information is used to reduce the appearance of motion blur in the images, and thereby improve diagnostic accuracy. The proposed personalized imaging approach builds on our group's extensive experience in machine learning. We expect that the combination of these various strategies will greatly reduce radiation dose, and because the improvements are based on software, the cost of these enhancements is very low. The project will result in recommendations for image reconstruction algorithms and parameters to be used in the clinic, along with corresponding tracer dose recommendations. The personalized imaging method will be implemented as a user-friendly computer program that customizes the dose for each given patient.
Radiation exposure of patients via medical imaging has become a significant concern. This project will investigate software methods that may allow the radiation dose in cardiac SPECT imaging to be reduced by a factor of eight or more. This will be accomplished by: optimizing the image processing that is used, further developing new and better image processing methods, and developing a method of 'personalized imaging', in which, for each individual patient, the minimum amount of imaging agent needed to obtain a good image is computed.
