Geraint Rees

Young-onset Alzheimer's Disease(AD) is a rare form of AD characterized by early symptom onset (< 65 years) and heterogeneous clinical phenotypes. Previous studies have consistently shown that patients with late-onset AD exhibit alterations in the default mode network-a large-scale brain network associated with self-related processing and autobiographical memory. However, the functional organization of the default-mode network is far less clear in young-onset AD. Here, we assessed default-mode network effective connectivity in two common young-onset AD variants (i.e., typical amnestic variant and posterior cortical atrophy) and healthy participants to identify disease- and variant-specific differences in the default-mode network. This case-control study was conducted with thirty-nine young-onset AD patients, including typical amnestic (n = 26, 15 females, mean age = 61) and posterior cortical atrophy (n = 13; 8 females, mean age = 61.8), and 24 age-matched healthy participants (13 females, mean age=60.1). All participants underwent resting-state functional MRI and extensive neuropsychological testing. Spectral dynamic causal modelling was performed to quantify resting-state effective connectivity between default-mode network regions. Parametric empirical Bayes analysis was then performed to characterise group differences in effective connectivity. Our results showed that patients with typical AD variant showed increased connectivity from medial prefrontal cortex to posterior default-mode network nodes as well as reduced inhibitory connectivity from hippocampus to other default-mode network nodes, relative to healthy controls …

The gold standard in the treatment of ischaemic stroke is set by evidence from randomised controlled trials. Yet the manifest complexities of the brain's functional, connective and vascular architectures introduce heterogeneity in treatment susceptibility that violates the premises of the underlying statistical framework, plausibly leading to substantial errors at both individual and population levels. The counterfactual nature of therapeutic inference has made quantifying the impact of this defect difficult. Employing large-scale lesion, connective, functional, genetic expression, and receptor distribution data, here we conduct a comprehensive series of semi-synthetic virtual interventional trials, quantifying the fidelity of the traditional approach in inferring individual treatment effects against biologically plausible, empirically informed ground truths. We compare the performance of machine learning models flexible enough to capture the observed heterogeneity, and find that the richness of the modelled lesion representation is decisive in determining individual-level fidelity, even where freedom from treatment allocation bias cannot be guaranteed. We are compelled to conclude that complex modelling of richly represented data is critical to individualised prescriptive inference in ischaemic stroke.

BackgroundBiotechnological syndromes refer to the illnesses that arise at the intersection of human physiology and digital technology. Now that we experience health and illness through so much technology (e.g. wearables, telemedicine, implanted devices), the medium is redefining our expression of symptoms, the observable signs of pathology and the range of diseases that may occur. Here, we systematically review all case reports describing illnesses related to digital technology in the past ten years, in order to identify novel biotechnological syndromes, map out new causal pathways of disease, and identify gaps in care that have disadvantaged a community of patients suffering from these digital complaints.MethodsPubMed, MEDLINE, Scopus, Cochrane Library and Web of Science were searched for case reports and case series that described patient cases involving biotechnological syndromes from 01/01 …

BackgroundThe prediction of long-term mortality following acute illness can be unreliable for older patients, inhibiting the delivery of targeted clinical interventions. The difficulty plausibly arises from the complex, multifactorial nature of the underlying biology in this population, which flexible, multimodal models based on machine learning may overcome. Here, we test this hypothesis by quantifying the comparative predictive fidelity of such models in a large consecutive sample of older patients acutely admitted to hospital and characterise their biological support.MethodsA set of 804 admission episodes involving 616 unique patients with a mean age of 84.5 years consecutively admitted to the Acute Geriatric service at University College Hospital were identified, in whom clinical diagnoses, blood tests, cognitive status, computed tomography of the head, and mortality within 600 days after admission were available. We …

Our knowledge of the organisation of the human brain at the population-level is yet to translate into power to predict functional differences at the individual-level, limiting clinical applications, and casting doubt on the generalisability of inferred mechanisms. It remains unknown whether the difficulty arises from the absence of individuating biological patterns within the brain, or from limited power to access them with the models and compute at our disposal. Here we comprehensively investigate the resolvability of such patterns with data and compute at unprecedented scale. Across 23810 unique participants from UK Biobank, we systematically evaluate the predictability of 25 individual biological characteristics, from all available combinations of structural and functional neuroimaging data. Over 4526 GPU*hours of computation, we train, optimize, and evaluate out-of-sample 700 individual predictive models, including multilayer perceptrons of demographic, psychological, serological, chronic morbidity, and functional connectivity characteristics, and both uni- and multi-modal 3D convolutional neural network models of macro- and micro-structural brain imaging. We find a marked discrepancy between the high predictability of sex (balanced accuracy 99.7%), age (mean absolute error 2.048 years, R2 0.859), and weight (mean absolute error 2.609Kg, R2 0.625), for which we set new state-of-the-art performance, and the surprisingly low predictability of other characteristics. Neither structural nor functional imaging predicted individual psychology better than the coincidence of common chronic morbidity (p<0.05). Serology predicted common morbidity (p …

Study on how the visual brain groups fragmented stimuli into coherent shapes & the role of stimulus awareness.

The quantification of cognitive powers rests on identifying a behavioural task that depends on them. Such dependence cannot be assured, for the powers a task invokes cannot be experimentally controlled or constrained a priori, resulting in unknown vulnerability to failure of specificity and generalisability. Evaluating a compact version of Raven's Advanced Progressive Matrices (RAPM), a widely used clinical test of fluid intelligence, we show that LaMa, a self-supervised artificial neural network trained solely on the completion of partially masked images of natural environmental scenes, achieves human-level test scores a prima vista, without any task-specific inductive bias or training. Compared with cohorts of healthy and focally lesioned participants, LaMa exhibits human-like variation with item difficulty, and produces errors characteristic of right frontal lobe damage under degradation of its ability to integrate global spatial patterns. LaMa's narrow training and limited capacity -- comparable to the nervous system of the fruit fly -- suggest RAPM may be open to computationally simple solutions that need not necessarily invoke abstract reasoning.

The visual cortex contains information about stimuli even when they are not consciously perceived. However, it remains unknown whether the visual system integrates local features into global objects without awareness. Here, we tested this by measuring brain activity in human observers viewing fragmented shapes that were either visible or rendered invisible by fast counterphase flicker. We then projected measured neural responses to these stimuli back into visual space. Visible stimuli caused robust responses reflecting the positions of their component fragments. Their neural representations also strongly resembled one another regardless of local features. By contrast, representations of invisible stimuli differed from one another and, crucially, also from visible stimuli. Our results demonstrate that even the early visual cortex encodes unconscious visual information differently from conscious information …

Achromatopsia is a congenital genetic condition of retinal cone dysfunction. While the rod signalling system in individuals with achromatopsia (ACHM) remains intact, the lack of cone signal leaves them with poor visual acuity, a foveal scotoma and complete colour blindness. We recently showed that gene therapies can restore cone function between eye and cortex in some patients, with individual differences in treatment effect. What may explain these differences? Previous studies have reported thickening of foveal visual cortex in ACHM and reduced overall visual cortex area, similar to changes observed in fully blind individuals. This may reflect deprivation related structural processes that can limit treatment amenability. Additionally, one study suggested that, in 3 individuals with ACHM, cortex taking input from the impaired fovea responds to rod-mediated inputs from the more peripheral retina, so treatment may …

Treatments are prescribed to individuals in pursuit of contemporaneously unobserved outcomes, based on evidence derived from populations with historically observed treatments and outcomes. Since neither treatments nor outcomes are typically replicable in the same individual, alternatives remain counterfactual in both settings. Prescriptive fidelity therefore cannot be evaluated empirically at the individual-level, forcing reliance on lossy, group-level estimates, such as average treatment effects, that presume an implausibly low ceiling on individuation. The lack of empirical ground truths critically impedes the development of individualised prescriptive models, on which realising personalised care inevitably depends. Here we present InterSynth, a general platform for modelling biologically-plausible, empirically-informed, semi-synthetic ground truths, for the evaluation of prescriptive models operating at the …

Background Whole‐brain longitudinal diffusion studies are crucial to examine changes in structural connectivity in neurodegeneration. Here, we investigated the longitudinal alterations in white matter (WM) microstructure across the timecourse of Huntington's disease (HD). Methods We examined changes in WM microstructure from premanifest to early manifest disease, using data from two cohorts with different disease burden. The TrackOn‐HD study included 67 controls, 67 premanifest, and 10 early manifest HD (baseline and 24‐month data); the PADDINGTON study included 33 controls and 49 early manifest HD (baseline and 15‐month data). Longitudinal changes in fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity, and radial diffusivity from baseline to last study visit were investigated for each cohort using tract‐based spatial statistics. An optimized pipeline was employed to generate …

Background Apathy, a disabling and poorly understood neuropsychiatric symptom, is characterised by impaired self-initiated behaviour. It has been hypothesised that the opportunity cost of time (OCT) may be a key computational variable linking self-initiated behaviour with motivational status. OCT represents the amount of reward which is foregone per second if no action is taken. Using a novel behavioural task and computational modelling, we investigated the relationship between OCT, self-initiation and apathy. We predicted that higher OCT would engender shorter action latencies, and that individuals with greater sensitivity to OCT would have higher behavioural apathy. Methods We modulated the OCT in a novel task called the ‘Fisherman Game’, Participants freely chose when to self-initiate actions to either collect rewards, or on occasion, to complete non-rewarding actions. We measured the relationship …

Healthcare policy, clinical practice and clinical research all declare patient benefit as their avowed aim. Yet, the conceptual question of what exactly constitutes patient benefit has received much less attention than the practical means of realising it. Currently, three key areas of conceptual unclarity make the achieved, real-world impact hard to quantify and disconnect it from the magnitude of the practical endeavour:(1) the distinction between objective and subjective benefit,(2) the relation between individual and population measures of benefit, and (3) the optimal measurement of benefit in research studies. A philosophical understanding of wellbeing is required to clarify these problems. Adopting a rigorous philosophical framework makes apparent that the differing goals of clinicians, researchers and research funders may make differing conceptions of patient benefit appropriate. A framework is proposed for …

Any clinically-deployed image-processing pipeline must be robust to the full range of inputs it may be presented with. One popular approach to this challenge is to develop predictive models that can provide a measure of their uncertainty. Another approach is to use generative modelling to quantify the likelihood of inputs. Inputs with a low enough likelihood are deemed to be out-of-distribution and are not presented to the downstream predictive model. In this work, we evaluate several approaches to segmentation with uncertainty for the task of segmenting bleeds in 3D CT of the head. We show that these models can fail catastrophically when operating in the far out-of-distribution domain, often providing predictions that are both highly confident and wrong. We propose to instead perform out-of-distribution detection using the Latent Transformer Model: a VQ-GAN is used to provide a highly compressed latent …

We describe CounterSynth, a conditional generative model of diffeomorphic deformations that induce label-driven, biologically plausible changes in volumetric brain images. The model is intended to synthesise counterfactual training data augmentations for downstream discriminative modelling tasks where fidelity is limited by data imbalance, distributional instability, confounding, or underspecification, and exhibits inequitable performance across distinct subpopulations.Focusing on demographic attributes, we evaluate the quality of synthesised counterfactuals with voxel-based morphometry, classification and regression of the conditioning attributes, and the Fréchet inception distance. Examining downstream discriminative performance in the context of engineered demographic imbalance and confounding, we use UK Biobank and OASIS magnetic resonance imaging data to benchmark CounterSynth …

Cortical cell loss is a core feature of Huntington’s disease (HD), beginning many years before clinical motor diagnosis, during the premanifest stage. However, it is unclear how genetic topography relates to cortical cell loss. Here, we explore the biological processes and cell types underlying this relationship and validate these using cell-specific post-mortem data. Eighty premanifest participants on average 15 years from disease onset and 71 controls were included. Using volumetric and diffusion MRI we extracted HD-specific whole brain maps where lower grey matter volume and higher grey matter mean diffusivity, relative to controls, were used as proxies of cortical cell loss. These maps were combined with gene expression data from the Allen Human Brain Atlas (AHBA) to investigate the biological processes relating genetic topography and cortical cell loss. Cortical cell loss was positively …

Background The HD Young Adult Study provided deep phenotyping of the earliest cohort of adult HD mutation carriers to date. We previously reported elevated levels of neurofilament light protein, suggestive of early neurodegenerative change. Putaminal volumes were also significantly reduced in HD mutation carriers. However, there was no evidence of any motor, cognitive or neuropsychiatric impairment approximately 24 years before expected disease onset.Aims We aim to follow up this valuable cohort to quantify longitudinal changes in imaging and biofluid markers and document premanifest emergence of motor, cognitive and neuropsychiatric changes.Methods We will undertake two follow up visits on our cohort of 64 young adult premanifest HD mutation carriers (preHD) and 67 matched controls at 5 and 6.5 years post baseline. In addition to the clinical, cognitive, neuropsychiatric, imaging and biofluid …

BackgroundConventional 3-Tesla MRI shows volume loss in the motor cortex (M1) up to 15 years before motor onset. Post-mortem data, typically at the end stage of Huntington's disease, suggests selective pyramidal cell loss of layers 3 and 5 in M1. However, the extent and patterns of cortical layer specific M1 pathology in the premanifest stage are unknown.AimsHere we present a preliminary data from the CLEAR-HD study. We demonstrate quantitative measurements of M1 cortical layers at 600μm resolution using MRI contrasts, R1, which is sensitive to myelin and R2*, which is sensitive to both myelin and iron.Methods7-Tesla MRI was used to acquire multi-parametric maps at 600μm resolution in 11 premanifest HD individuals and 14 healthy controls. The HCP-MMP 1.0 atlas was used to parcellate the motor cortex and R1 and R2* values were sampled across 8 equi-volume cortical layers, generated using …

Correctly estimating the influence of uncertainty on our decisions is a critical metacognitive faculty. However, the relationship between sensory uncertainty (or its inverse, precision), decision accuracy, and subjective confidence is currently unclear. Although some findings indicate that healthy adults exhibit an illusion of over-confidence, under-confidence in response to sensory uncertainty has also been reported. One reason for this ambiguity is that stimulus intensity and precision are typically confounded with one another, limiting the ability to assess their independent contribution to metacognitive biases. Here we report four psychophysical experiments controlling these factors, finding that healthy human participants are systematically under-confident when discriminating low-precision stimuli. This bias remains even when decision accuracy and reaction time are accounted for, indicating that a performance-independent computation partially underpins the influence of sensory precision on confidence. We further show that this influence is linked to fluctuations in arousal and individual differences in the neuroanatomy of the left superior parietal lobe and middle insula. These results illuminate the neural and physiological correlates of precision misperception in metacognition.Significance StatementThe ability to recognize the influence of sensory uncertainty on our decisions underpins the veracity of self-monitoring, or metacognition. In the extreme, a systematic confidence bias can undermine decision accuracy and potentially underpin disordered self-insight in neuropsychiatric illness. Previously it was unclear if metacognition accurately reflects …

The characteristics and determinants of health and disease are often organized in space, reflecting our spatially extended nature. Understanding the influence of such factors requires models capable of capturing spatial relations. Drawing on statistical parametric mapping, a framework for topological inference well established in the realm of neuroimaging, we propose and validate an approach to the spatial analysis of diverse clinical data—GeoSPM—based on differential geometry and random field theory. We evaluate GeoSPM across an extensive array of synthetic simulations encompassing diverse spatial relationships, sampling, and corruption by noise, and demonstrate its application on large-scale data from UK Biobank. GeoSPM is readily interpretable, can be implemented with ease by non-specialists, enables flexible modeling of complex spatial relations, exhibits robustness to noise and under-sampling …