Imiomics and Deep Learning MRI and PET-MRI Studies on Causes and Consequences of Body Composition in Cardiovascular Disease

Bakgrund: We are facing a global epidemic of obesity and related cardiovascular complications. Understanding of the underlying mechanisms is key for development of novel intervention strategies. The associations between body composition (adipose tissue and muscle content), risk factors for cardiovacular disease (CVD) (e.g. diabetes type 2, atherosclerosis) and cardiovascular endpoints as consequences of vulnerable plaques (cardiac and brain infarcts) are not fully understood.

The understanding of obesity biology has been greatly assisted by the advances within the field of genetics. Imiomics is an automated whole-body MRI and PET/MRI analysis concept, based on image registration that deforms all data to a common coordinate system, which allows for each voxel intensity (e.g. fat content and metabolic activity) to be compared between individuals and within an individual over time, and also to be correlated with non-imaging data (e.g. genotypes and disease phenotypes).

Deep learning with convolutional neural networks dominates the state-of-the-art in image analysis tasks.

Målsättning: Our intention is to improve our knowledge about how obesity, including the total amount of body fat and its distribution, causes diabetes type 2 and atherosclerosis with associated cardiovascular endpoints. We also want to study which genes and proteins that regulate body composition and its consequences.

Arbetsplan: Aiming for these goals, we will continue to apply Imiomics and Deep Learning methods for automated analyses of whole-body MRI and PET/MRI, integrating also non-imaging data, relevant for CVD and its risk factors, from several well-characterised cohorts. The workplan can be divided into four work packages:

1) Apply Imiomics to build a normal whole-body Human Imaging Atlas from large-scale cohort studies (n>70.000) and to investigate sex- and age-related differences in body composition 2) Find how genetic variants, relevant for CVD, affect body composition 3) Find whole-body body composition associations with CVD (e.g. ischemic stroke, myocardial infarction) and its risk factors (e.g. type 2 diabetes, hypertension, dyslipidemia) 4) Find causal effects of body composition estimated by instrumental variable analysis

Betydelse: By this approach, new important findings for future prevention and therapy strategies of CVD and its risk factors are anticipated to be revealed, despite potentially being difficult or impossible to detect with more conventional methods.