CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Bayesian methods in biomedical imaging Authors:  Michele Guindani - University of California, Irvine (United States) [presenting]
Abstract: The use of flexible Bayesian approaches in biomedical imaging is discussed. We will first discuss applications to the analysis of task-related fMRI data in single-subject and multi-subject experiments, where the aim is to account for the heterogeneity in neuronal activity both within- and between- subjects. We will then discuss an application to cancer radiomics, an emerging discipline that promises to elucidate lesion phenotypes and tumor heterogeneity through the analysis of large amounts of quantitative imaging features that can be derived from medical images. We will show how a fully Bayesian probabilistic framework may help characterizing the heterogeneity of adrenal lesions images obtained from CT scans more precisely than a class of machine-learning approaches currently used in cancer radiomics. We further assess whether the subtypes resultant from our analysis are clinically oriented by investigating their correspondence with pathological diagnoses.