Algorithms for anatomy-guided image reconstruction in nuclear medicine tomography.

PET and SPECT images are more often combined with anatomical information from CT or MRI, but usually only after the image formation. The image quality of the emission images can be improved by using the anatomical images during the PET or SPECT reconstruction. In this project, multiple anatomy-guided MAP algorithms are investigated. Previously, the A-MAP algorithm, which was developed by our research group, has been validated for FDG-PET and MRI of epilepsy patients. In this project, A-MAP will be adapted and evaluated for other tracers and other cortical disorders. A-MAP requires an accurate segmentation of the anatomy. Good segmentation software is available for the cortical brain structures. Subcortical structures are more difficult, and also for "whole body" CT images segmentation is a problem. Therefore, MAp algorithms without segmentation will be developed. These algorithms will be evaluated for PET/MRI imaging of neurological disorders in which subcortical structures are involved, and for PAT/CT imaging in oncology and radiotherapy. For this, it is important that the anatomical and PET images are aligned perfectly. Non-rigid registration algorithms will be studied as a means to correct for erroneous alignment, e.g., due to patient movement.

Keywords:Image quality, Nuclear medicine, Anatomy, Neurology, Oncology, Radiotherapy

Disciplines:Medical imaging and therapy, Other paramedical sciences, Multimedia processing, Biological system engineering, Signal processing