Mosaics of polynomial transformations giving a patient specific registration to reduce breathing motion artefacts

Nuclear Medicine (NM) imaging is an important diagnostic tool, used widely in the field of oncology amongst others. However, NM images suffer from significant blurring due to inevitable patient motion, such as breathing, coughing and other voluntary or involuntary movement. Advances in detector technology and reconstruction techniques have led to a steady improvement in NM spatial resolution and the problems posed by patient motion are therefore becoming increasingly significant, particularly for PET/CT. Many correction schemes will require gated images to be aligned. This paper describes a method of registering major organs in piecewise fashion called virtual dissection. Results from processing synthetic data and one set of patient data are presented. A key feature is the possibility to use a mixture of registration techniques on a single set of data and combine the results into a single set of output images.