Respiratory motion correction in PET with super-resolution techniques and non-rigid registration

Respiratory motion causes image degradation and artifacts in positron emission tomography (PET) images and constitutes an important source or error for the interpretation and proper quantization of thoracic and abdominal studies. Current work towards the correction of respiratory motion effects is mostly based on the acquisition of respiratory synchronized data which, on the other hand, leads to images with low statistics and therefore increased noise level. Super- Resolution (SR) techniques deal with how to combine several images into an enhanced high-resolution image, and in this work we study the application of SR to respiratory-gated PET images to produce a respiratory motion compensated image that incorporates the information provided by all the gated frames acquired. A B-spline based non-rigid registration algorithm is used for estimating the deformations among frames. We present preliminary results on a simulated PET/CT image sequence, corresponding to a thoracic study where a number of different sized nodular lesions were included inside the lungs. The resulting PET image allows better discrimination of the lesions and improved contrast, compared to each of the gated frames and to simple averaging of all the frames after registration.