Relationship between brain volumes and cardiac image derived phenotypes

Abstract

Introduction It has been extensively shown that the insurgence of dementia and cognitive decline is strongly tied to the presence of stroke and brain vascular damage. However, while hypertension is a major risk factor for stroke, the relationship between cardiac abnormalities such as high blood pressure or arrhythmias, with dementia still needs to be elucidated. Currently, large biobank studies offer the chance of jointly analyzing the common variation between cardiovascular pathologies and neurodegeneration. Methods From the UK Biobank database, we have image information from 10,000 individuals. From these, we have selected 4424 participants, which had all the complete image information, cardiac and brain images. For each individual, we extracted cardiovascular and brain related phenotypes. The cardiovascular information includes : systolic blood pressure, diastolic blood pressure, cardiac output and stroke volume. As brain volumetric indicators, we took into account grey matter (GM) and white matter (WM) volumes, and the ventricles volume. All the volumes were normalized by the total head volume. The relationship among the different variables was studied through canonical correlation analysis (CCA). Results and discussion Fig. 1 shows the comparison between the canonical loadings obtained from CCA, as well as the loadings obtained when regressing out age from the model. When age is not taken into account, the main predictor of the correlation is the systolic blood pressure, followed by diastolic blood pressure. These variables are correlated with the shrinkage of the brain volume (WM and GM) and with the increase of ventricles volume. When age is explicitly accounted for, we observe that cardiac output and stroke volume become the dominant factors, being mostly associated with GM loss and ventricles volumes increase. We hypothesize that in the first scenario age was strongly implicitly associated with blood pressures and brain changes. After removing the influence of age, we can observe how GM decreases as cardiac output and stroke volume increase, while WM shrinkage has disappeared. This indicates an underlying relationship among cardiovascular indicators and brain volumes, which should be further explored. We plan to explore this relationship in the future, developing unsupervised multivariate methods and biophysical models that allow us to obtain additional features that can help us to characterize this behavior and hopefully identify pathological subjects.