Prof. Nils Gunnar Wahlgren (Chair), Department of Clinical Neuroscience Karolinska Institute, Sweden. Dr. Myzoon Ali (VISTA Coordinator) University of Glasgow www.vistacollaboration.org vista.plus@glasgow.ac.uk Semi-automated holistic measures of brain health using magnetic resonance imaging after stroke Proposed Investigators: Dr David Alexander Dickie, Prof Jesse Dawson, Dr Francesco Arba, Prof Joanna Wardlaw.

A large number of stroke survivors either have, or go on to develop, cognitive decline, dementia, or recurrent stroke[1]. We are interested in whether we can better predict these outcomes through novel analysis of data contained within brain MRI scans. These analysis techniques use data contained in routine clinical sequences and do not require additional or ‘novel’ MR sequences. Upwards of 90% of stroke patients show background brain changes that are thought to represent cerebral small vessel disease (SVD) and many have evidence of atrophy. These features are known to be associated with worse outcomes. For example, the presence of white matter hyperintensities (WMH) is associated with risk of recurrent stroke, cognitive deficits, and gait problems[1]. Other features such as presence of cerebral microbleeds, enlarged perivascular spaces and silent lacunar infarcts are also inversely associated with cognitive function[2]. However, these relationships, while statistically significant, typically explain only a fraction of the variance in outcome and cognition (r~0.10)[3, 4]. We know that combinations of brain MRI data strengthen these associations. For example, summed visual rating scales for WMH, lacunes, microbleeds, and PVS scores are more closely associated with cardiovascular risk factors and cognitive outcomes in community-dwelling volunteers and patients with minor stroke[2, 5]. We want to explore whether true quantitative measures of these MRI features, combined with additional measures such as atrophy, provide greater predictive power. Dr Dickie and Prof Wardlaw have developed a semi-automated brain health measure, the “Brain Health Index” (under review in European Radiology). This measure incorporates information from T1, T2, FLAIR, and T2* MRI into a Gaussian cluster analysis. BHI explained more than twice the variance in Addenbrooke’s Cognitive Exam (ACER) than summed SVD scores and WMH volume in 150 people with mild stroke. We are looking to further develop this method in VISTA data, through optimising the algorithm with fewer sequences, and assessing generalisability in different scanners and a different cohort of stroke patients. We already know that there is a relationship between cognitive impairment and visual rating scales for SVD in VISTA (OR=1.72, 95% CI=1.22-2.42) and we want to test whether these relationships can be strengthened with BHI.

We hypothesize that an automated whole brain damage metric (BHI) will have greater predictive power for cognitive function and recurrent stroke than individual assessments of markers of SVD and brain atrophy.

We will usedata from VISTA including patients with previous stroke who had an MRI scan performed, a measure of cognitive function, and details of recurrent stroke events. We request all available MRI sequences and require at least a T1 and FLAIR image for each patient. Additionally requested variables are listed in the table, where available. We will use multiple linear regression to account for confounders between brain imaging metrics and outcome data (column 2 in Table). Finally, we request the visual rating scales from included MRI scans where available (from the study by led by Francesco Arba).

We aim to better predict cognitive and vascular outcomes after stroke. This could help guide selection in clinical trials but also guide treatment (should effective treatments for cognitive decline after stroke be found).

We have funds to pay for access to VISTA data.