Brain network analysis for neuroimaging PET data
The brain is increasingly recognized to operate as an integrated system where multiple regions work together to perform complex tasks. In this context, the importance of methods that can highlight patterns of neurochemical activity (as opposed to localized behavior), either in a single brain system or as interactions between systems, is becoming increasingly relevant. These observations warrant the implementation of novel approaches to the analysis of positron emission tomography (PET) imaging data, capable to (i) capture the network-like properties of neurochemical systems; (ii) extract joint and unique information between several functionally related neurochemical systems and (iii) extract biologically meaningful features that are most sensitive to disease progression and mechanisms. This talk will highlight how methods such as multiset Canonical Correlation Analysis (mCCA) and Dynamic Mode Decomposition (DMD) applied to high quality quantitative multi-tracer PET data extend the investigative power of this imaging modality and provide new insights into neurodegenerative processes.
Advances in PET/MR multimodality imaging: relevance to the study of brain function
Recent understanding of brain function stresses the importance of the interaction between brain connectivity and underlying neurochemistry and metabolism, both in terms of energy cost of brain function as well as in terms of understanding of pathogenic processes. Indeed, the network degeneration hypothesis states that initiation and progression of disease-specific pathological changes occur within specific brain structural and functional networks (best investigated with MRI) and are mediated by abnormal protein aggregation, inflammation and impaired cellular energetics coupled to abnormal neurotransmission (best investigated with PET). In order to best study brain function in the above described context it is important that PET and MRI imaging is performed simultaneously, that the time and spatial resolutions of PET and MRI-based imaging are optimally matched and that relevant information can be extracted from multi-parameter data by identifying task-specific most informative combinations of imaging metrics. This talk will describe the development in PET and MRI imaging techniques that was spurred by these requirements, including improvements in imaging instrumentation/performance, introduction of novel image reconstruction approaches, image denoising, data and process modeling, as well as application of data fusion and machine learning approaches to data analysis.
Dr. Vesna Sossi received the Laurea degree from the University of Trieste, Italy, in High Energy Physics and the PhD degree from the University of British Columbia, Vancouver, B.C., Canada, in Nuclear Physics. She is a Faculty member in the UBC Physics and Astronomy Department and leads UBC PET imaging. She has first worked on detectors and data analysis as applied to measurements of nuclear reactions cross sections at the Canadian Nuclear Physics Laboratory TRIUMF and then transitioned to Nuclear Medicine. Since then, she has worked in many areas ranging from instrumentation related topics such as development of data reconstruction and quantification algorithms, design and development of a preclinical MR-compatible PET insert, to more applied areas such as development of novel analysis methods for multi-tracer PET data, which led to new insights into neurodegeneration. She has published more than 200 papers and 280 abstracts; she sits on several national and international review committees and is a reviewer for many journals and conferences. She has been attending the IEEE MIC meetings since 1993 and has served on Nuclear Medical and imaging Sciences Council (NMISC), the NPSS AdCom, was MIC Program chair in 2012 and NSS/MIC General Chair in 2015 and 2023.