Low-count-high-quality reconstructions for PET and SPECT

Image noise and dose given to the patient are of critical importance in numerous clinical applications of nuclear imaging methods (Single Photon Emission Computer Tomography SPECT, Positron Emission Tomography - PET). For example, high image noise levels represent an important drawback for obtaining an optimal clinical outcome from PET or SPECT scans in cardiology. Likewise, for paediatric oncology imaging and for imaging patients with neurological and psychiatric disorders, the amount of radioactive tracer (and resulting dose) administered to the patient is of concern, too. The aim of this project is to improve the image quality of SPECT and PET images at very low counting statistics by developing new image reconstruction techniques. The proposed methods, which are based on advanced image reconstruction techniques, will allow obtaining high-quality images from shortened PET and SPECT examinations or examinations requiring lower amounts of injected activity levels, both of which contributing to increased patient comfort. Monte Carlo simulations of numerical phantoms and experimental phantom acquisitions will be used for the validation of the proposed reconstruction algorithms. Furthermore, the proposed image reconstruction methods will be evaluated in PET and SPECT acquisitions of pilot patient data from clinical routine. The novel image reconstruction approaches developed within this project are expected to help improving the clinical diagnostics of patients with paediatric tumours or patients with neurological, psychiatric or cardiovascular diseases.

The collaboration project between the Medical University of Vienna (MUV) and the Novosibirsk Technical University (NTU) focusing on the development of a new image reconstruction algorithm using local regularization. The project started in October 2017 and lasted for 3 years with a 4th year being added without renumeration. During this project, one PhD candidate was hired to work with the MUV team (and another PhD position being filled by the NTU team). The NTU team did focus on the theoretical foundations of the new reconstruction methodology, while the Austrian team did engage in the development of new simulation models for SPECT and PET imaging and reconstruction validation. Following the joint funding, all project partner activities were well aligned. The first image reconstructions of phantoms demonstrated promising results in terms of improved contrast-to-noise ratios of phantom data. However, prior to clinical adoption, more investigations - also with clinical data - are required. The project also included an evaluation of new PET image reconstruction methods applied to non-lesional epilepsy patients at low count levels; work done in collaboration with the KU Leuven (scientific advisor to the project). The main project also led to a spin-off study to investigate the potential of reducing the injected activity levels in paediatric patients. Both studies demonstrated clinical feasibility with the latest PET imaging system technology (aka higher volume sensitivity). In summary, the collaboration partners were satisfied with the current outcome and are prepared to take this joint research further, by automating the choice of the regularization factor in the image reconstruction and validating this type of reconstruction with more, clinically-relevant data.