Jose Bernal

Brain maintenance – the preservation of brain structure or function relevant to cognitive performance – remains challenging to quantify. Here, we propose a domain-general brain maintenance index derived by jointly modelling the longitudinal co-evolution of ageing-related atrophy (via medial temporal lobe to ventricle ratio, MTLV-ratio), white matter hyperintensities (WMH), and global cognition assessed by the preclinical Alzheimer’s cognitive composite (PACC5) using latent growth curve modelling. We demonstrate its utility in 543 cognitively unimpaired older adults from the DELCODE cohort, followed annually over four years. We show that changes in MTLV-ratio and WMH additively predict cognitive change. We further show that higher neuroticism, depressive symptoms, lower openness, and faster biological ageing are related to unfavourable domain-specific trajectories and poorer brain maintenance. Our findings highlight the combined relevance of WMH and ageing-related atrophy dynamics for brain maintenance. Maintaining cerebrovascular and mental health alongside cognitive engagement could promote brain maintenance, delay cognitive decline and dementia.

Background: Blood-brain barrier (BBB) dysfunction is increasingly recognized as a feature of cerebral amyloid angiopathy (CAA) and has been linked to hemorrhagic imaging manifestations such as cortical superficial siderosis. However, it remains unclear whether neurovascular barrier dysfunction can be captured by routinely available fluid biomarkers and whether such markers identify clinically relevant hemorrhage-prone CAA phenotypes. The CSF/serum albumin quotient (QAlb) is an established marker of neurovascular barrier dysfunction. We investigated QAlb levels in CAA and their association with imaging markers of disease severity. Methods: We included 225 participants (115 with CAA, 72 with Alzheimers disease [AD], 38 healthy controls) with CSF biomarkers and standardized MRI evaluation. Pathologic QAlb levels were identified via the age-corrected Reiber-formula. Group differences and determinants of pathological QAlb were assessed using uni- and multivariable regression analyses. The diagnostic relevance was assessed by receiver operating characteristic analysis. Results: QAlb levels were higher in CAA than in controls (ratio of means [RoM] 1.43, 95% CI 1.28-1.58) and patients with AD (RoM 1.22, 95% CI 1.10-1.35; both p<0.001). Pathological QAlb was independently associated with CAA compared with controls (OR 12.16, 95% CI 2.56-57.86) and AD (OR 2.14, 95% CI 1.07-4.28). Despite these associations, QAlb showed only moderate discrimination between CAA and controls (AUC 0.75, 95% CI 0.67-0.82) and low discrimination between CAA and AD (AUC 0.63, 95% CI 0.55-0.71). Within the CAA cohort, pathological QAlb was independently associated with disseminated cortical superficial siderosis (OR 3.92, 95% CI 1.31-11.72; p=0.014), a marker of advanced hemorrhage-prone disease. Conclusion: QAlb is elevated in patients with CAA and is associated with disseminated cortical superficial siderosis, the strongest predictor of future intracerebral hemorrhage. These findings support an association between neurovascular barrier dysfunction and the hemorrhage-prone phenotype of CAA.

BACKGROUND: Centrum semiovale perivascular spaces (CSO PVS) are common to different cerebral small vessel disease (CSVD) subtypes, yet their enlargement to the point of visibility on magnetic resonance imaging is thought to occur through distinct mechanisms. In cerebral amyloid angiopathy (CAA), CSO PVS are thought to reflect impaired perivascular Aβ (amyloid-β) drainage, whereas in deep perforator arteriopathy (DPA) they may result from microangiopathy-related arterial stiffening and altered fluid dynamics. The extent to which this can be confirmed in vivo remains unclear. METHODS: We retrospectively analyzed 186 patients with CSVD and cerebrospinal fluid (CSF) Aβ biomarkers (n=111 probable CAA, n=75 DPA). CSO PVS were counted on axial T2-weighted images. Associations between CSO PVS and CSF biomarkers were assessed via Pearson correlation and multivariable linear regression, including an interaction term between CSF Aβ and CSVD subtype, adjusted for demographics and neuroimaging markers of CSVD. RESULTS: Patients with higher CSO PVS counts were generally younger, had lower white matter hyperintensity burden, higher basal ganglia PVS counts, and were more frequently affected by cortical superficial siderosis. CSO PVS counts were similar in patients with CAA and DPA. The association between CSF Aβ _42/40 ratio and CSO PVS burden was observed in patients with CAA, but not in those with DPA (interaction term between CSF Aβ _42/40 ratio and CAA: β=–0.27; P =0.016), independent of demographics and other neuroimaging markers of CSVD. CSF Aβ ₄₀ showed no association with CSO PVS counts in any model. CONCLUSIONS: Our findings strengthen the pathophysiological link between CSO PVS and Aβ pathology in CAA but not in DPA. These results extend previous histopathologic and neuroimaging work and underscore the need to interpret CSO PVS in the context of underlying CSVD subtype.

Chronic psychosocial stress (CPS) is associated with adverse brain and mental health outcomes. Effects on the cerebral microvasculature have been proposed as an underlying mechanism, although this remains to be established. Here, we examined the association between CPS and an early marker of microvascular dysfunction, magnetic resonance imaging (MRI)-visible perivascular spaces (PVS). Analyses were conducted in two cohorts of healthy young adults (N = 61; ages 18-43 years; 88% male) using high-resolution 3T MRI and an automated PVS quantification pipeline. CPS was assessed using the Perceived Stress Scale (PSS-10). We applied a two-step meta-analytic framework and controlled for known allostatic factors impacting PVS, including age, body mass index and mean arterial pressure. In accordance with our hypothesis, individuals with higher CPS had significantly higher fractional PVS volumes in the centrum semiovale (CSO), in particular in the frontal and occipital lobes (p _FDR < .05). No such effect was found in the basal ganglia, or in the CSO subdivision, parietal, and temporal lobes (p _FDR > .09). Our findings indicate that CPS may contribute to subtle, centrum semiovale specific microvascular alterations even in healthy young adults. Future multimodal research including inflammatory marker and blood-brain barrier measures may help to elucidate mechanistic pathways.

Background: Cognitive reserve (CR) and brain maintenance (BM) support preserved cognition by mitigating or slowing neural decline in aging. Physical fitness represents a plausible pathway to both CR and BM, yet the underlying neurobiological mechanisms remain insufficiently understood. This study examined the role of physical fitness in BM and CR by testing its associations with Alzheimer's disease related brain pathology and by assessing its moderating and mediating effects on the relationship between pathology and cognition. Methods: Data were obtained from 345 cognitively unimpaired older adults (mean age = 73.11, SD 8.03 years; 177 females) participating in the ongoing SFB1436 study. We assessed global cognition and delayed verbal memory performance, aerobic fitness (VO2max), and muscular capacity. Blood biomarkers included plasma Aβ1-42/Aβ1-40, p-tau217, and GFAP (glial fibrillary acidic protein), as well as serum BDNF (brain derived neurotrophic factor), VEGF (vascular endothelial growth factor), and Cathepsin-B (plasticity). Neuroimaging measures comprised medial temporal lobe (MTL) tau burden ([18F]PI-2620 PET), MRI-derived hippocampal volumes, white matter hyperintensities, perivascular spaces (PVS) in basal ganglia (BG) and centrum semiovale, total gray matter volume (GMV), and MTL thickness. To examine BM, we tested associations between fitness and brain pathology. To examine CR, we conducted moderation analyses assessing whether fitness attenuated the negative impact of MTL tau burden on cognition. Mediation analyses further evaluated whether hippocampal volume, total GMV, or MTL thickness mediated a potential association between fitness and cognition. All models controlled for age and sex (and education in mediation analyses), and multiple comparisons were FDR-corrected. Results: Physical fitness was not related to cognitive performance. Higher muscular capacity was associated with lower BG PVS volumes, while higher aerobic fitness was related to higher MTL thickness and total GMV. Elevated MTL tau burden was associated with poorer verbal memory performance. Although physical fitness did not significantly moderate the tau-memory relationship, model comparison indicated weak evidence for CR effects (p = .015). Conclusion: Muscular capacity was linked to lower BG PVS volumes, supporting resistance against pathology, while aerobic fitness related to preserved cortical integrity and tended to act as a CR proxy against MTL tau pathology. Physical fitness may support cerebrovascular and glymphatic function, providing weak CR-like protection against tau-related cognitive decline.

Background: MRI-visible perivascular spaces (PVS) are an imaging marker of compromised brain health function. However, robust PVS segmentation remains challenging: conventional methods often require cohort-specific parameter tuning, while supervised deep-learning models depend on extensive manual annotations for training and may fail to generalise to unseen scanners, protocols, and clinical populations. Methods: We introduce DRIPS (Domain Randomisation for Image-based PVS Segmentation), a fully automated deep-learning framework trained exclusively on synthetic data. By learning PVS representations from a plethora of simulated cases, DRIPS transfers to real-world MRI without using manually labelled PVS data from real participants. We evaluated DRIPS in 185 fully manually annotated MRI scans from eight independent cohorts encompassing diverse clinical populations and acquisition settings, including 3T and 7T imaging, T1-weighted and T2-weighted sequences, and isotropic and anisotropic resolutions. DRIPS was compared with five previously published methods. Results: Across cohorts, DRIPS significantly outperformed the second-best method for both area under the precision-recall curve (AUPRC) and centreline Dice similarity coefficient (clDice), with mean gains of at least 0.1: 0.146 for basal ganglia (BG) AUPRC (P = 0.008), 0.125 for centrum semiovale (CSO) AUPRC (P = 0.008), 0.112 for BG clDice (P = 0.016), and 0.186 for CSO clDice (P = 0.008). Importantly, its performance was not associated with white matter hyperintensity burden. Conclusions: DRIPS enables accurate, fully automated PVS segmentation across heterogeneous imaging settings without the need for real PVS annotations for training.

Background: Cerebral small vessel disease (CSVD) is a common neurological condition that contributes to strokes, dementia, disability, and mortality worldwide. We conducted a systematic review and meta-analysis to investigate the use of neuroimaging CSVD markers in machine learning (ML) based diagnosis and prognosis of cognitive impairment and dementia, and identify both methodological changes over time and barriers to clinical translation. Methods: Following the PRISMA guidelines, we systematically searched for original studies that used both neuroimaging CSVD markers and ML methods for diagnosing and prognosing neurodegenerative diseases (preregistration in PROSPERO: CRD42022366767). Each paper was independently reviewed by a pair of reviewers at all stages, with a third consulted to resolve conflicts. We meta-analysed the effectiveness of ML models to distinguish healthy controls from Alzheimer’s dementia and cognitive impairment, using area under the curve (AUC) as the performance metric. Results: We identified 75 studies: 43 on diagnosis, 27 on prognosis, and 5 on both. Nearly 60% of studies were published in the past two years, reflecting a growing interest in using CSVD markers in ML-based diagnosis and prognosis of neurodegenerative diseases, especially Alzheimer’s dementia. This rising interest may be linked to the strong performance of such models: according to our meta-analysis, ML approaches using CSVD markers perform well in differentiating healthy controls from Alzheimer’s dementia (AUC 0.88 [95%-CI 0.85–0.92]) and cognitive impairment (AUC 0.84 [95%-CI 0.74–0.95]). However, the growing interest has not been matched by methodological rigour: only 16 studies met the criteria for inclusion in the meta-analysis due to inconsistent reporting, only five assessed the generalisability of their models on external datasets, and six lacked clear diagnostic criteria. Conclusions: Interest in incorporating CSVD markers into ML models for neurodegenerative disease classification is on the rise, and their performance suggests that this is worth further exploration. Serious methodological issues, including inconsistent reporting, limited generalisability testing, and other potential biases, are unfortunately common and hinder further adoption. Our targeted recommendations provide a roadmap to accelerate the integration of ML into clinical practice.