Search

Background: The sources of inter‐ and intra‐individual variability in age‐related cognitive decline remain poorly understood. We examined the association between 20‐year trajectories of cognitive decline and multimodal brain structure and morphology in older age. Method: We used the Whitehall II Study, an extensively characterised cohort with 3T brain magnetic resonance images acquired at older age (mean age = 69.52 ± 4.9) and 5 repeated cognitive performance assessments between mid‐life (mean age = 53.2 ±4.9 years) and late‐life (mean age = 67.7 ± 4.9). Using non‐negative matrix factorisation, we identified 10 brain components integrating cortical thickness, surface area, fractional anisotropy, and mean and radial diffusivities (Figure 1). Result: We observed two latent variables describing distinct brain‐cognition associations. The first describes variations in 5 structural components associated with low mid‐life performance across multiple cognitive domains, decline in reasoning, but maintenance of fluency abilities. The second describes variations in 6 structural components associated with low mid‐life performance in fluency and memory, but retention of multiple abilities (Figure 2). Expression of latent variables predicts future cognition 3.2 years later (mean age = 70.87 ± 4.9) (Figure 3). Conclusion: Longitudinal cognitive decline and maintenance across diverse cognitive functions are both positively and negatively associated with distinct markers of cortical structure. Latent brain‐behaviour relationships predict future cognitive performance. [ABSTRACT FROM AUTHOR]

Background: The co‐occurrence of multiple long term conditions (LTCs), termed multimorbidity, is a risk factor for dementia. However, we lack an understanding of how and when specific LTCs accumulate over time to impact dementia risk. Method: 338,270 individuals from the UK Biobank with linked electronic health records up to January 2022 were studied, excluding those with either a dementia diagnosis or age of latest follow up < 65 years. LTC diagnoses were derived using hospital episode statistics and GP records (age of diagnosis computed using month/year of birth and month/year of diagnosis). We performed an age‐stratified analysis in which we a) cluster individuals based on 43 non‐dementia LTCs diagnosed within age bins (0‐55, 55‐65, 65‐70 years), and b) compute risk of incident dementia using logistic regression with multimorbidity cluster as predictor and incident dementia as outcome. Diagnosis data was submitted to PCA, and the resulting factor scores were kMeans‐clustered. 6 clusters were derived per age bin. Result: We named clusters based on the most prevalent LTCs within that cluster (Figure 1). Cardiometabolic, arthritis, gastrointestinal, and mixed clusters existed at all ages. Mental health and respiratory clusters were present exclusively in the younger age bin (0‐55) whereas hypertension and eye disorder clusters were exclusive to older ages (55‐65 and 65‐70). Across the lifespan, compared to being disease free, multimorbidity was significantly associated with incident dementia (OR>1). We also observed age‐specific multimorbidity preceding dementia onset. From 0‐55 years, being in the mixed cluster presented the highest dementia risk (OR = 5.13, 95% CI = [4.61, 5.73]). Being in the cardiometabolic cluster was associated with highest risk in 55‐65 (OR = 6.24 [5.48, 7.1]) and 65‐70 (OR = 6.99 [5.91, 8.27]) ranges. The most predictive trajectory of incident dementia involved the mixed cluster at 55, followed by the cardiometabolic clusters from 55‐65 and 65‐70. Table 1 displays the top 10 trajectories associated with dementia. All p‐values are bonferroni‐holmes corrected and < 0.01. Conclusion: We present multimorbidity pathways linked to increased risk of dementia. We describe how and when diseases cluster together prior to dementia diagnosis. [ABSTRACT FROM AUTHOR]

Background: Cerebrovascular biomarkers such as cerebrovascular reactivity (CVR) have previously been associated with episodic memory. However, the relationship between memory and regional CVR within the hippocampus is not well studied. Moreover, whether this association is affected by hippocampal volume remains to be investigated. This study investigated the associations between hippocampal physiology, structure, and memory in ageing. Method: Data from 104 participants (Age: Mean = 77±4.9 years) of the ongoing Oxford Heart and Brain Study was analysed for this study. CVR was assessed by a compensatory change in the blood oxygenation level dependent magnetic resonance imaging (MRI) signal in response to a 5% CO2 challenge, hippocampus volume was measured by structural MRI, and episodic memory was assessed with the Hopkins Verbal Learning Test Revised (HVLT‐R). Result: Mean hippocampal CVR decreased with age (β = ‐0.004, p = 0.006). Mean hippocampal CVR was positively associated with mean hippocampal volume (β = 0.11, p = 0.010), and this association was driven by the left hippocampus after correcting for age and sex. Higher hippocampal CVR was also associated with better performance on the HVLT (β = 13.8, p = 0.035). This association was fully mediated by hippocampal volume, suggesting a path from cerebrovascular reactivity to episodic memory (Sobel's z = 2.16, p = 0.030, see Figure 1). Conclusion: This study presents novel findings showing that hippocampal CVR was associated with episodic memory, and this association was mediated by hippocampal volume. This suggests a pathway by which impairments in hippocampal CVR are associated with worse episodic memory via smaller hippocampal volumes. [ABSTRACT FROM AUTHOR]

Background: Disturbance of iron homeostasis in the brain is linked to cognitive impairment and neurodegenerative diseases. Interactions between changes of iron loads in the blood and changes in the brain are poorly understood. No strong and complete understanding of structural changes in the brain associated with changing values of blood measurements such as haemoglobin concentration has been established. We aimed to establish a link between changes in blood measurements and the accompanying structural changes in the brain. Method: We analysed 4436 participants from Visit 2 with a mean age of 61.92 years (standard deviation = 7.42). T2* image derived phenotypes, age, sex and blood count to represent iron were used in a linear correlation analysis. We used age, sex, blood count, head volumes, scanning positions and differences between T1 and swMRI for correction in a linear model approach predicting swMRI voxel values. We applied linear correlation analysis and a general linear model using all extracted information. Further, we built a convolutional neural network (CNN) to compare approaches. The CNN used swMRI as input and blood values as output. The blood values were grouped based on percentiles to create classes. Attention maps were extracted from the CNN using Integrated Gradients to identify possible regions driving the decision process of the network for the correct classifications. Result: No linear relation between blood count and brain imaging was detected in the Visit 2 participant subset we analysed. However, we achieved an accuracy of over 0.99 for predicting haemoglobin concentration values split into 3, 10 or 20 classes for our CNN model. The attention maps showed high attention in the thalamus and the parietal, occipital and temporal lobe. Conclusion: We were able to predict blood iron using MRI imaging. This suggests an association between the two measures. Furthermore, it was possible to show which brain areas were used by the CNN for decision‐making. Using such an integrated model of iron modalities may give a more complete view of iron homeostasis and the potential associations with related diseases such as dementia. [ABSTRACT FROM AUTHOR]

Background: Disturbance of iron homeostasis in the brain is linked to cognitive impairment and neurodegenerative diseases. Interactions between changes of iron loads in the blood and changes in the brain are poorly understood. No strong and complete understanding of structural changes in the brain associated with changing values of blood measurements such as haemoglobin concentration has been established. We aimed to establish a link between changes in blood measurements and the accompanying structural changes in the brain. Method: We analysed 4436 participants from Visit 2 with a mean age of 61.92 years (standard deviation = 7.42). T2* image derived phenotypes, age, sex and blood count to represent iron were used in a linear correlation analysis. We used age, sex, blood count, head volumes, scanning positions and differences between T1 and swMRI for correction in a linear model approach predicting swMRI voxel values. We applied linear correlation analysis and a general linear model using all extracted information. Further, we built a convolutional neural network (CNN) to compare approaches. The CNN used swMRI as input and blood values as output. The blood values were grouped based on percentiles to create classes. Attention maps were extracted from the CNN using Integrated Gradients to identify possible regions driving the decision process of the network for the correct classifications. Result: No linear relation between blood count and brain imaging was detected in the Visit 2 participant subset we analysed. However, we achieved an accuracy of over 0.99 for predicting haemoglobin concentration values split into 3, 10 or 20 classes for our CNN model. The attention maps showed high attention in the thalamus and the parietal, occipital and temporal lobe. Conclusion: We were able to predict blood iron using MRI imaging. This suggests an association between the two measures. Furthermore, it was possible to show which brain areas were used by the CNN for decision‐making. Using such an integrated model of iron modalities may give a more complete view of iron homeostasis and the potential associations with related diseases such as dementia. [ABSTRACT FROM AUTHOR]

Background: Although several volumetric studies have attempted to examine the relationship between hippocampal subfields and age, their findings are heterogenous. Here, in addition to typical volumetric measures, we investigated age‐related alterations of hippocampal subfield shape and microstructure. Method: The present work included 161 healthy participants (ages 18‐81) recruited as part of two local datasets. Standard T1‐weighted (T1w; MPRAGE sequence, 1 mm3 voxels) and high‐resolution T2‐weighted (T2w, 0.64 mm3 voxels) magnetic resonance images were acquired. The minc‐bpipe‐library pipeline (https://github.com/CobraLab/minc‐bpipe‐library) was used to preprocess images. The MAGeT‐Brain algorithm (Pipitone et al., 2014) was utilized for segmentation of the hippocampal subfields (Amaral et al., 2016). The volume and shape (i.e. surface area [SA] and displacement) of the grey matter (GM) subfields (i.e. Cornus Ammonis [CA]1, CA2CA3, CA4 and dentate gyrus [CA4DG], stratum radiatum/lacunosum/moleculare and subiculum) and the volume of associated white matter (WM) regions (i.e. fimbria, fornix, alveus and mammillary bodies [MB]) were estimated using T1w images. Mean T1w/T2w signal intensity ratio (Glasser & Van Essen, 2011) of each subfield was calculated to examine microstructural alterations sensitive to myelin. Linear‐mixed effects models and Akaike information criterion were used to select the most appropriate age relationships from multiple fits (i.e. linear, quadratic, and cubic) for each structure of interest. Sex and ipsilateral hippocampal GM or WM volume were included as covariates. Statistical analyses were performed in R 3.5.0; comparisons surviving p<0.05 (Bonferroni‐corrected) were reported. Result: Right CA1 and alveus volumes were positively cubically related to age and bilateral CA4DG volumes were negatively linearly related to age. T1w/T2w intensity linearly decreased with age for all subregions except for bilateral fornix and MB, which expressed quadratic evolution with age (Figures 1‐2). Age was also inversely related to bilateral SA reduction in the head and medial‐posterior hippocampus, and lateral outward and medial inward displacement (Figure 3). Conclusion: The current work suggests that certain subfields of the hippocampus are more (e.g. CA1) or less (e.g. CA4DG) preserved in healthy aging. These differences do not appear to be explained by microstructural variation. [ABSTRACT FROM AUTHOR]

Background: Globally, up to 40% of dementia cases may be prevented if several key risk factors, such as low education and obesity, are targeted. This has motivated interest into the development of risk scores that aim to quantify an individual's risk of developing dementia within a given time frame. However, translation to a clinical setting has been hampered by either a lack of external validation, poor out‐of‐sample performance, or the integration of measures (e.g., MRI, cognitive testing) that are costly or time intensive to administer. This study aimed to develop a novel dementia risk score suitable for a primary care setting and compare its discrimination accuracy with other risk scores. Method: 208,108 UK Biobank (UKB) participants were examined, with cases of incident dementia (mean time to diagnosis ± SD = 5.91 years ± 2.02) identified through self‐report, hospital inpatient records and death certificates (n = 1,270, 0.6%). The dataset was randomly split into a training (80%) and test set (20%). Potential predictors (n = 30) included Apolipoprotein E4 (APOE4) status and cardiovascular, medical and lifestyle factors collected at baseline. Logistic LASSO regression was employed for variable selection and logistic regression was then used to derive beta coefficients for each predictor included in the UKB dementia risk score (UKB‐DRS). The area under the curve (i.e., AUC) was used to compare the discriminative ability of the UKB‐DRS to the CAIDE, ANU‐ADRI, Dementia Risk Score (i.e., DRS) and Framingham risk score, in both the test set and the Whitehall II study (WHIII, n = 2,974), which served as an external dataset. Result: The UKB‐DRS consisted of age, sex, education, APOE4 status, a medical history of diabetes, stroke and depression and a family history of dementia. The UKB‐DRS demonstrated good discrimination accuracy in both the test set (AUC [95% ] = 0.79 [0.77, 0.82]) and external dataset (AUC [95% CI] = 0.83 [0.79,0.87]). Additionally, the UKB‐DRS outperformed all other risk scores (the DRS had the next best AUC [95% CI]: UKB = 0.77 [0.74, 0.8]; WHII = 0.77 [0.73, 0.81]). Conclusion: This study contributes a novel dementia risk score that may be used to guide primary prevention strategies. [ABSTRACT FROM AUTHOR]