MOTION CORRECTION OF FREELY MOVING BODIES IN POSITRON EMISSION TOMOGRAPHY

Abstract

Positron emission tomography (PET) scan is a non-invasive nuclear medical imaging technique that produces a 3D picture of the functional processes of the body. During this 1─2 hour preclinical research procedure which uses positron-emitting radioisotopes, the patient is normally restricted in terms of its body movements. It becomes very difficult to restrain an infant or a patient in pain. Under such situations, the patients are usually injected with anesthetic drugs which are known to interfere with key biological processes such as neural-hemodynamic coupling and receptor binding. The ability to image a freely moving body would not only help overcome the above mentioned current technological limitations but would also represent a critical step towards understanding certain phenomenon such as understanding the relationship between brain function and behavior, etc. A free movement of the body under observation using conventional high resolution PET scanner can result into false registration of positron emitted events (i.e., as if events were originated from a false location). It would make the image reconstruction process almost impossible by using the conventional methods. Moreover, the data acquisition in the case of a freely moving body is further complicated by a complex attenuation field and scattering problems. Since access to a million dollar PET scanner machine is very limited in Pakistan, this work is focused towards investigating the feasibility of imaging a freely moving digital phantom (imitating the patient body) by using a virtual PET scanner. The motion of the object is tracked by using one or more than one cameras mounted on the patient bed and processing its output images either by the same computer that is reconstructing the acquired data or by another computer (hardware dependency).To imitate a real life detection and reconstruction process, more than one open source software were used. XCAT was used to create a custom phantom representing the subject under observation. ASIM (Analytical PET simulator) was used to simulate the emission of positrons from the said body. Siemens ECAT 962 was the real life PET scanner that was mimicked in the simulation. For Image reconstruction, both MATLAB and viii STIR were used while the AMIDE and Image J were used to manipulate the input and output slices. Current work is restricted to some basic motion as the target of study was to introduce a sound principle that can be further implemented to more complex movements that the patient may have. To detect the patient movement, the proposed technique uses images from cameras mounted on the PET machine. Any motion detected is compensated during the sinogram creation of event data, hence eradicating most of the expected scatter and attenuation. Reconstructed images are further refined by image registration techniques. This leads to a correlation coefficient of up to 0.85 between the phantom slice and the reconstructed images. All explored techniques are implemented and their results are compared. The demos and tutorial of all the above mentioned open source programs are added in Annex C to help future participants have a better understanding of current work.