UK Biobank imaging in a memory clinic population: associations between imaging‐derived phenotypes and cognition in the Oxford Brain Health Clinic

Abstract

AbstractBackgroundSince August 2020, the Oxford Brain Health Clinic (BHC) has seen over 200 NHS memory clinic patients (O’Donoghue et al., 2022). In addition to high‐quality cognitive and lifestyle assessments and opportunities for research participation, patients receive a clinical MRI scan (T1‐weighted, T2‐FLAIR, and SWI) and can consent to additional research scans (diffusion MRI – dMRI, resting‐state functional MRI – rfMRI, and arterial spin labelling ‐ ASL) aligned to the UK Biobank (Griffanti et al., 2022). In this project, we aimed to automatically extract imaging‐derived phenotypes (IDPs) from the different imaging modalities acquired at the BHC and perform unimodal group‐level analyses to explore associations with cognition and diagnoses in this real‐world memory clinic population.MethodAs of January 2023, scans were analysed from 176 BHC patients, 101 of whom completed the additional research MRI scans. Cognitive scores (ACE‐III total score, N = 166) were available from the BHC appointment, and subsequent diagnoses (N = 118) were extracted from electronic healthcare records. Scans were processed using the UK Biobank pipeline (Alfaro‐Almagro et al., 2018), adapted to include previously described modifications (Griffanti et al., 2022). All IDPs were deconfounded for age, sex, and head size. We calculated Spearman correlations with ACE‐III cognitive scores and group comparisons (Kruskal‐Wallis tests) between 3 diagnostic groups: dementia, MCI, and no dementia‐related diagnosis. Results were corrected for multiple comparisons (false discovery rate with Benjamini‐Hochberg procedure; FDR‐adjusted p = 0.000016).Result11 IDPs significantly correlated with ACE‐III (Figure 1). Four voxel‐based morphometry temporal lobe measures, 3 FreeSurfer temporal lobe measures, FIRST left hippocampal volume, and SIENAX peripheral grey matter volume positively correlated with ACE‐III, in line with known dementia‐related atrophy patterns (Figure 2). dMRI mean diffusivity in the bilateral parahippocampal part of the cingulum negatively correlated with ACE‐III, highlighting white matter disruption (Figure 3). No FDR‐corrected differences were found between diagnostic groups.ConclusionUsing an unselected patient population, all of whom have a degree of memory problems, this work provides real‐world validation of associations that are well‐established in the research context. This represents a key step towards integrating research‐quality imaging in the memory clinic. Imaging and clinical variables are planned to be available through DPUK.