Model-based imaging and image-based modelling
Daniel Alexander, University College London
Tuesday March 23, 12-1pm
Zoom Link: https:/mcgill.zoom.us/j/91589192037
Abstract: My talk will cover three current research topics. 1. Microstructure imaging, which uses mathematical or computational modelling and machine learning to estimate and map microstructural features of tissue from MRI; see (Alexander NMR Biomed 2019) for a review. Examples from my group’s work include NODDI (Zhang et al Neuroimage 2012) for brain imaging and VERDICT (Panagiotaki et al Cancer Research 2014) for cancer imaging. I will describe moves towards a new paradigm combining multi-contrast measurements through sophisticated computational models exploiting machine learning. 2. Data-driven disease progression models (e.g. Fonteijn et al Neuroimage 2012; Young et al Brain 2014; Lorenzi et al Neuroimage 2017), which aim to piece together longitudinal pictures of disease from cross-sectional or short-term longitudinal data sets and thus gain new disease understanding, stratification systems, and predictive power. The Subtype and Stage Inference (SuStaIn - Young et al Nature Comms 2018) algorithm extends the idea to identify disease subgroups defined by distinct longitudinal trajectories of change. Most recently we are exploiting these models to make inferences about underlying spreading mechanisms in neurodegeneration (Garbarino et al eLife 2019). 3. Image Quality Transfer (Alexander et al Neuroimage 2017; Tanno et al Neuroimage 2020), which uses machine learning to estimate a high quality image, e.g. that we would have acquired from a one-off super-powered scanner, from a lower quality image, e.g. acquired on a standard hospital scanner.