Development of a Neural Network-Based Model for Generating CT uMaps from Standalone PET Imaging

One type of diagnostic imaging used in nuclear medicine is positron emission tomography, (PET). PET images display the function of the tissue being imaged, for example sugar metabolism, without displaying the anatomy. In hospitals, people undergoing PET imaging will also be imaged with computed tomography (CT) which displays the anatomy, and the two images are overlapped.
The extra CT image is time consuming and provides the person being imaged with a small extra radiation dose. Cubresa is working to develop a standalone PET imaging system for faster image collection and a lower radiation dose at the cost of reduced anatomical information. One issue with images that rely on radiation, such as PET and CT, is that some of the radiation that comes from the person being image does not make it to the detectors to make the image because of a process called attenuation. Extensive research has been done and CT images are corrected for attenuation. Similar research is not as thoroughly completed with PET images because the attenuation corrections from the CT images can be used on the PET images. This project aims to begin further development of the attenuation correct for standalone PET images.

Faculty Supervisor:

Melanie Martin

Student:

Partner:

Cubresa Inc

Discipline:

Physics

Sector:

Manufacturing; Professional, scientific and technical services

University:

University of Winnipeg

Program: