Computational limits to the legibility of the imaged human brain.

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 23 810 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 fully-connected feed-forward neural networks of demographic, psychological, serological, chronic disease, 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 an individual's psychology better than the coincidence of common chronic disease (p < 0.05). Serology predicted chronic disease (p < 0.05) and was best predicted by it (p < 0.001), followed by structural neuroimaging (p < 0.05). Our findings suggest either more informative imaging or more powerful models will be needed to decipher individual level characteristics from the human brain. We make our models and code openly available.

Compressed representation of brain genetic transcription.

The architecture of the brain is too complex to be intuitively surveyable without the use of compressed representations that project its variation into a compact, navigable space. The task is especially challenging with high-dimensional data, such as gene expression, where the joint complexity of anatomical and transcriptional patterns demands maximum compression. The established practice is to use standard principal component analysis (PCA), whose computational felicity is offset by limited expressivity, especially at great compression ratios. Employing whole-brain, voxel-wise Allen Brain Atlas transcription data, here we systematically compare compressed representations based on the most widely supported linear and non-linear methods-PCA, kernel PCA, non-negative matrix factorisation (NMF), t-stochastic neighbour embedding (t-SNE), uniform manifold approximation and projection (UMAP), and deep auto-encoding-quantifying reconstruction fidelity, anatomical coherence, and predictive utility across signalling, microstructural, and metabolic targets, drawn from large-scale open-source MRI and PET data. We show that deep auto-encoders yield superior representations across all metrics of performance and target domains, supporting their use as the reference standard for representing transcription patterns in the human brain.

Comprehensive Assessment of Nutrition and Dietary Influences in Hypermobile Ehlers-Danlos Syndrome-A Cross-Sectional Study.

INTRODUCTION: Disorders of gut-brain interaction (DGBI) are common in patients with hypermobile Ehlers-Danlos syndrome/hypermobility spectrum disorder (hEDS/HSD). Food is a known trigger for DGBI symptoms, which often leads to dietary alterations and, increasingly, nutrition support. We aimed to explore dietary behaviors and influencing factors in patients with hEDS/HSD. METHODS: In a cross-sectional study, patients with hEDS/HSD were recruited from Ehlers-Danlos Support UK (nontertiary) and tertiary neurogastroenterology clinics to complete questionnaires characterizing the following: dietary behaviors, nutrition support, DGBI (Rome IV), gastrointestinal symptoms, anxiety, depression, avoidant restrictive food intake disorder (ARFID), mast cell activation syndrome, postural tachycardia syndrome (PoTS), and quality of life. We used stepwise logistic regression to ascertain which factors were associated with dietary behaviors and nutrition support. RESULTS: Of 680 participants (95% female, median age 39 years), 62.1% altered their diet in the last year and 62.3% regularly skipped meals. Altered diet was associated with the following: reflux symptoms ( P < 0.001), functional dyspepsia ( P = 0.008), reported mast cell activation syndrome ( P < 0.001), and a positive screen for ARFID, specifically fear of eating and low interest ( P < 0.001). Approximately 31.7% of those who altered their diet required nutrition support. The strongest predictor of requiring nutrition support was a positive screen for ARFID, specifically fear of eating (OR: 4.97, 95% CI: 2.09-11.8, P < 0.001). DISCUSSION: Altered diet is very common in the patients with hEDS/HSD we studied and influenced by functional dyspepsia, reflux symptoms, and ARFID. Those with ARFID have a 4-fold increased risk of requiring nutrition support, and therefore, it is paramount that psychological support is offered in parallel with dietary support in the management of DGBI in hEDS/HSD.

Graph lesion-deficit mapping of fluid intelligence.

Fluid intelligence is arguably the defining feature of human cognition. Yet the nature of its relationship with the brain remains a contentious topic. Influential proposals drawing primarily on functional imaging data have implicated 'multiple demand' frontoparietal and more widely distributed cortical networks, but extant lesion-deficit studies with greater causal power are almost all small, methodologically constrained, and inconclusive. The task demands large samples of patients, comprehensive investigation of performance, fine-grained anatomical mapping, and robust lesion-deficit inference, yet to be brought to bear on it. We assessed 165 healthy controls and 227 frontal or non-frontal patients with unilateral brain lesions on the best-established test of fluid intelligence, Raven's Advanced Progressive Matrices, employing an array of lesion-deficit inferential models responsive to the potentially distributed nature of fluid intelligence. Non-parametric Bayesian stochastic block models were used to reveal the community structure of lesion deficit networks, disentangling functional from confounding pathological distributed effects. Impaired performance was confined to patients with frontal lesions [F(2,387) = 18.491; P < 0.001; frontal worse than non-frontal and healthy participants P < 0.01, P <0.001], more marked on the right than left [F(4,385) = 12.237; P < 0.001; right worse than left and healthy participants P < 0.01, P < 0.001]. Patients with non-frontal lesions were indistinguishable from controls and showed no modulation by laterality. Neither the presence nor the extent of multiple demand network involvement affected performance. Both conventional network-based statistics and non-parametric Bayesian stochastic block modelling heavily implicated the right frontal lobe. Crucially, this localization was confirmed on explicitly disentangling functional from pathology-driven effects within a layered stochastic block model, prominently highlighting a right frontal network involving middle and inferior frontal gyrus, pre- and post-central gyri, with a weak contribution from right superior parietal lobule. Similar results were obtained with standard lesion-deficit analyses. Our study represents the first large-scale investigation of the distributed neural substrates of fluid intelligence in the focally injured brain. Combining novel graph-based lesion-deficit mapping with detailed investigation of cognitive performance in a large sample of patients provides crucial information about the neural basis of intelligence. Our findings indicate that a set of predominantly right frontal regions, rather than a more widely distributed network, is critical to the high-level functions involved in fluid intelligence. Further they suggest that Raven's Advanced Progressive Matrices is a useful clinical index of fluid intelligence and a sensitive marker of right frontal lobe dysfunction.

Brain tumour genetic network signatures of survival.

Tumour heterogeneity is increasingly recognized as a major obstacle to therapeutic success across neuro-oncology. Gliomas are characterized by distinct combinations of genetic and epigenetic alterations, resulting in complex interactions across multiple molecular pathways. Predicting disease evolution and prescribing individually optimal treatment requires statistical models complex enough to capture the intricate (epi)genetic structure underpinning oncogenesis. Here, we formalize this task as the inference of distinct patterns of connectivity within hierarchical latent representations of genetic networks. Evaluating multi-institutional clinical, genetic and outcome data from 4023 glioma patients over 14 years, across 12 countries, we employ Bayesian generative stochastic block modelling to reveal a hierarchical network structure of tumour genetics spanning molecularly confirmed glioblastoma, IDH-wildtype; oligodendroglioma, IDH-mutant and 1p/19q codeleted; and astrocytoma, IDH-mutant. Our findings illuminate the complex dependence between features across the genetic landscape of brain tumours and show that generative network models reveal distinct signatures of survival with better prognostic fidelity than current gold standard diagnostic categories.

Constipation Predominant Irritable Bowel Syndrome and Functional Constipation Are Not Discrete Disorders: A Machine Learning Approach.

INTRODUCTION: Chronic constipation is classified into 2 main syndromes, irritable bowel syndrome with constipation (IBS-C) and functional constipation (FC), on the assumption that they differ along multiple clinical characteristics and are plausibly of distinct pathophysiology. Our aim was to test this assumption by applying machine learning to a large prospective cohort of comprehensively phenotyped patients with constipation. METHODS: Demographics, validated symptom and quality of life questionnaires, clinical examination findings, stool transit, and diagnosis were collected in 768 patients with chronic constipation from a tertiary center. We used machine learning to compare the accuracy of diagnostic models for IBS-C and FC based on single differentiating features such as abdominal pain (a "unisymptomatic" model) vs multiple features encompassing a range of symptoms, examination findings and investigations (a "syndromic" model) to assess the grounds for the syndromic segregation of IBS-C and FC in a statistically formalized way. RESULTS: Unisymptomatic models of abdominal pain distinguished between IBS-C and FC cohorts near perfectly (area under the curve 0.97). Syndromic models did not significantly increase diagnostic accuracy (P > 0.15). Furthermore, syndromic models from which abdominal pain was omitted performed at chance-level (area under the curve 0.56). Statistical clustering of clinical characteristics showed no structure relatable to diagnosis, but a syndromic segregation of 18 features differentiating patients by impact of constipation on daily life. DISCUSSION: IBS-C and FC differ only about the presence of abdominal pain, arguably a self-fulfilling difference given that abdominal pain inherently distinguishes the 2 in current diagnostic criteria. This suggests that they are not distinct syndromes but a single syndrome varying along one clinical dimension. An alternative syndromic segregation is identified, which needs evaluation in community-based cohorts. These results have implications for patient recruitment into clinical trials, future disease classifications, and management guidelines.

Functional brain networks and neuroanatomy underpinning nausea severity can predict nausea susceptibility using machine learning.

KEY POINTS: Nausea is an adverse experience characterised by alterations in autonomic and cerebral function. Susceptibility to nausea is difficult to predict, but machine learning has yet to be applied to this field of study. The severity of nausea that individuals experience is related to the underlying morphology (shape) of the subcortex, namely of the amygdala, caudate and putamen; a functional brain network related to nausea severity was identified, which included the thalamus, cingulate cortices (anterior, mid- and posterior), caudate nucleus and nucleus accumbens. Sympathetic nervous system function and sympathovagal balance, by heart rate variability, was closely related to both this nausea-associated anatomical variation and the functional connectivity network, and machine learning accurately predicted susceptibility or resistance to nausea. These novel anatomical and functional brain biomarkers for nausea severity may permit objective identification of individuals susceptible to nausea, using artificial intelligence/machine learning; brain data may be useful to identify individuals more susceptible to nausea. ABSTRACT: Nausea is a highly individual and variable experience. The central processing of nausea remains poorly understood, although numerous influential factors have been proposed, including brain structure and function, as well as autonomic nervous system (ANS) activity. We investigated the role of these factors in nausea severity and if susceptibility to nausea could be predicted using machine learning. Twenty-eight healthy participants (15 males; mean age 24 years) underwent quantification of resting sympathetic and parasympathetic nervous system activity by heart rate variability. All were exposed to a 10-min motion-sickness video during fMRI. Neuroanatomical shape differences of the subcortex and functional brain networks associated with the severity of nausea were investigated. A machine learning neural network was trained to predict nausea susceptibility, or resistance, using resting ANS data and detected brain features. Increasing nausea scores positively correlated with shape variation of the left amygdala, right caudate and bilateral putamen (corrected P = 0.05). A functional brain network linked to increasing nausea severity was identified implicating the thalamus, anterior, middle and posterior cingulate cortices, caudate nucleus and nucleus accumbens (corrected P = 0.043). Both neuroanatomical differences and the functional nausea-brain network were closely related to sympathetic nervous system activity. Using these data, a machine learning model predicted susceptibility to nausea with an overall accuracy of 82.1%. Nausea severity relates to underlying subcortical morphology and a functional brain network; both measures are potential biomarkers in trials of anti-nausea therapies. The use of machine learning should be further investigated as an objective means to develop models predicting nausea susceptibility.

Artificial Intelligence-Assisted Gastroenterology- Promises and Pitfalls.

Technological advances in artificial intelligence (AI) represent an enticing opportunity to benefit gastroenterological practice. Moreover, AI, through machine or deep learning, permits the ability to develop predictive models from large datasets. Possibilities of predictive model development in machine learning are numerous dependent on the clinical question. For example, binary classifiers aim to stratify allocation to a categorical outcome, such as the presence or absence of a gastrointestinal disease. In addition, continuous variable fitting techniques can be used to predict quantity of a therapeutic response, thus offering a tool to predict which therapeutic intervention may be most beneficial to the given patient. Namely, this permits an important opportunity for personalization of medicine, including a movement from guideline-specific treatment algorithms to patient-specific ones, providing both clinician and patient the capacity for data-driven decision making. Furthermore, such analyses could predict the development of GI disease prior to the manifestation of symptoms, raising the possibility of prevention or pre-treatment. In addition, computer vision additionally provides an exciting opportunity in endoscopy to automatically detect lesions. In this review, we overview the recent developments in healthcare-based AI and machine learning and describe promises and pitfalls for its application to gastroenterology.

Morphology of subcortical brain nuclei is associated with autonomic function in healthy humans.

The autonomic nervous system (ANS) is a brain body interface which serves to maintain homeostasis by influencing a plethora of physiological processes, including metabolism, cardiorespiratory regulation and nociception. Accumulating evidence suggests that ANS function is disturbed in numerous prevalent clinical disorders, including irritable bowel syndrome and fibromyalgia. While the brain is a central hub for regulating autonomic function, the association between resting autonomic activity and subcortical morphology has not been comprehensively studied and thus was our aim. In 27 healthy subjects [14 male and 13 female; mean age 30 years (range 22-53 years)], we quantified resting ANS function using validated indices of cardiac sympathetic index (CSI) and parasympathetic cardiac vagal tone (CVT). High resolution structural magnetic resonance imaging scans were acquired, and differences in subcortical nuclei shape, that is, 'deformation', contingent on resting ANS activity were investigated. CSI positively correlated with outward deformation of the brainstem, right nucleus accumbens, right amygdala and bilateral pallidum (all thresholded to corrected P < 0.05). In contrast, parasympathetic CVT negatively correlated with inward deformation of the right amygdala and pallidum (all thresholded to corrected P < 0.05). Left and right putamen volume positively correlated with CVT (r = 0.62, P = 0.0047 and r = 0.59, P = 0.008, respectively), as did the brainstem (r = 0.46, P = 0.049). These data provide novel evidence that resting autonomic state is associated with differences in the shape and volume of subcortical nuclei. Thus, subcortical morphological brain differences in various disorders may partly be attributable to perturbation in autonomic function. Further work is warranted to investigate these findings in clinical populations. Hum Brain Mapp 39:381-392, 2018. © 2017 Wiley Periodicals, Inc.

The Role of Esophageal Hypersensitivity in Functional Esophageal Disorders.

The Rome IV diagnostic criteria delineates 5 functional esophageal disorders which include functional chest pain, functional heartburn, reflux hypersensitivity, globus, and functional dysphagia. These are a heterogenous group of disorders which, despite having characteristic symptom profiles attributable to esophageal pathology, fail to demonstrate any structural, motility or inflammatory abnormalities on standard clinical testing. These disorders are associated with a marked reduction in patient quality of life, not least considerable healthcare resources. Furthermore, the pathophysiology of these disorders is incompletely understood. In this narrative review we provide the reader with an introductory primer to the structure and function of esophageal perception, including nociception that forms the basis of the putative mechanisms that may give rise to symptoms in functional esophageal disorders. We also discuss the provocative techniques and outcome measures by which esophageal hypersensitivity can be established.

Molecular neurobiology of addiction: what's all the (Δ)FosB about?

The transcription factor ΔFosB is upregulated in numerous brain regions following repeated drug exposure. This induction is likely to, at least in part, be responsible for the mechanisms underlying addiction, a disorder in which the regulation of gene expression is thought to be essential. In this review, we describe and discuss the proposed role of ΔFosB as well as the implications of recent findings. The expression of ΔFosB displays variability dependent on the administered substance, showing region-specificity for different drug stimuli. This transcription factor is understood to act via interaction with Jun family proteins and the formation of activator protein-1 (AP-1) complexes. Once AP-1 complexes are formed, a multitude of molecular pathways are initiated, causing genetic, molecular and structural alterations. Many of these molecular changes identified are now directly linked to the physiological and behavioral changes observed following chronic drug exposure. In addition, ΔFosB induction is being considered as a biomarker for the evaluation of potential therapeutic interventions for addiction.

VASARI-auto: Equitable, efficient, and economical featurisation of glioma MRI.

The VASARI MRI feature set is a quantitative system designed to standardise glioma imaging descriptions. Though effective, deriving VASARI is time-consuming and seldom used clinically. We sought to resolve this problem with software automation and machine learning. Using glioma data from 1172 patients, we developed VASARI-auto, an automated labelling software applied to open-source lesion masks and an openly available tumour segmentation model. Consultant neuroradiologists independently quantified VASARI features in 100 held-out glioblastoma cases. We quantified 1) agreement across neuroradiologists and VASARI-auto, 2) software equity, 3) an economic workforce analysis, and 4) fidelity in predicting survival. Tumour segmentation was compatible with the current state of the art and equally performant regardless of age or sex. A modest inter-rater variability between in-house neuroradiologists was comparable to between neuroradiologists and VASARI-auto, with far higher agreement between VASARI-auto methods. The time for neuroradiologists to derive VASARI was substantially higher than VASARI-auto (mean time per case 317 vs. 3 s). A UK hospital workforce analysis forecast that three years of VASARI featurisation would demand 29,777 consultant neuroradiologist workforce hours and >£1.5 ($1.9) million, reducible to 332 hours of computing time (and £146 of power) with VASARI-auto. The best-performing survival model utilised VASARI-auto features instead of those derived by neuroradiologists. VASARI-auto is a highly efficient and equitable automated labelling system, a favourable economic profile if used as a decision support tool, and non-inferior survival prediction. Future work should iterate upon and integrate such tools to enhance patient care.

The minimal computational substrate of fluid intelligence.

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 representative 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 suggest matrix-style tests may be open to computationally simple solutions that need not necessarily invoke the substrates of reasoning.

The human cost of ethical artificial intelligence.

Foundational models such as ChatGPT critically depend on vast data scales the internet uniquely enables. This implies exposure to material varying widely in logical sense, factual fidelity, moral value, and even legal status. Whereas data scaling is a technical challenge, soluble with greater computational resource, complex semantic filtering cannot be performed reliably without human intervention: the self-supervision that makes foundational models possible at least in part presupposes the abilities they seek to acquire. This unavoidably introduces the need for large-scale human supervision-not just of training input but also model output-and imbues any model with subjectivity reflecting the beliefs of its creator. The pressure to minimize the cost of the former is in direct conflict with the pressure to maximise the quality of the latter. Moreover, it is unclear how complex semantics, especially in the realm of the moral, could ever be reduced to an objective function any machine could plausibly maximise. We suggest the development of foundational models necessitates urgent innovation in quantitative ethics and outline possible avenues for its realisation.

Brain tumour segmentation with incomplete imaging data.

Progress in neuro-oncology is increasingly recognized to be obstructed by the marked heterogeneity-genetic, pathological, and clinical-of brain tumours. If the treatment susceptibilities and outcomes of individual patients differ widely, determined by the interactions of many multimodal characteristics, then large-scale, fully-inclusive, richly phenotyped data-including imaging-will be needed to predict them at the individual level. Such data can realistically be acquired only in the routine clinical stream, where its quality is inevitably degraded by the constraints of real-world clinical care. Although contemporary machine learning could theoretically provide a solution to this task, especially in the domain of imaging, its ability to cope with realistic, incomplete, low-quality data is yet to be determined. In the largest and most comprehensive study of its kind, applying state-of-the-art brain tumour segmentation models to large scale, multi-site MRI data of 1251 individuals, here we quantify the comparative fidelity of automated segmentation models drawn from MR data replicating the various levels of completeness observed in real life. We demonstrate that models trained on incomplete data can segment lesions very well, often equivalently to those trained on the full completement of images, exhibiting Dice coefficients of 0.907 (single sequence) to 0.945 (complete set) for whole tumours and 0.701 (single sequence) to 0.891 (complete set) for component tissue types. This finding opens the door both to the application of segmentation models to large-scale historical data, for the purpose of building treatment and outcome predictive models, and their application to real-world clinical care. We further ascertain that segmentation models can accurately detect enhancing tumour in the absence of contrast-enhancing imaging, quantifying the burden of enhancing tumour with an R 2 > 0.97, varying negligibly with lesion morphology. Such models can quantify enhancing tumour without the administration of intravenous contrast, inviting a revision of the notion of tumour enhancement if the same information can be extracted without contrast-enhanced imaging. Our analysis includes validation on a heterogeneous, real-world 50 patient sample of brain tumour imaging acquired over the last 15 years at our tertiary centre, demonstrating maintained accuracy even on non-isotropic MRI acquisitions, or even on complex post-operative imaging with tumour recurrence. This work substantially extends the translational opportunity for quantitative analysis to clinical situations where the full complement of sequences is not available and potentially enables the characterization of contrast-enhanced regions where contrast administration is infeasible or undesirable.

A framework for focal and connectomic mapping of transiently disrupted brain function.

The distributed nature of the neural substrate, and the difficulty of establishing necessity from correlative data, combine to render the mapping of brain function a far harder task than it seems. Methods capable of combining connective anatomical information with focal disruption of function are needed to disambiguate local from global neural dependence, and critical from merely coincidental activity. Here we present a comprehensive framework for focal and connective spatial inference based on sparse disruptive data, and demonstrate its application in the context of transient direct electrical stimulation of the human medial frontal wall during the pre-surgical evaluation of patients with focal epilepsy. Our framework formalizes voxel-wise mass-univariate inference on sparsely sampled data within the statistical parametric mapping framework, encompassing the analysis of distributed maps defined by any criterion of connectivity. Applied to the medial frontal wall, this transient dysconnectome approach reveals marked discrepancies between local and distributed associations of major categories of motor and sensory behaviour, revealing differentiation by remote connectivity to which purely local analysis is blind. Our framework enables disruptive mapping of the human brain based on sparsely sampled data with minimal spatial assumptions, good statistical efficiency, flexible model formulation, and explicit comparison of local and distributed effects.

Brain effect of transcutaneous vagal nerve stimulation: A meta-analysis of neuroimaging evidence.

BACKGROUND: Dysfunction in the autonomic nervous system is common throughout many functional gastrointestinal diseases (FGIDs) that have been historically difficult to treat. In recent years, transcutaneous vagal nerve stimulation (tVNS) has shown promise for improving FGID symptoms. However, the brain effects of tVNS remain unclear, which we investigated by neuroimaging meta-analysis. METHODS: A total of 157 studies were identified, 4 of which were appropriate for inclusion, encompassing 60 healthy human participants. Using activation likelihood analysis estimation, we statistically quantified functional brain activity changes across three domains: (1) tVNS vs. null stimulation, (2) tVNS vs. sham stimulation, and (3) sham stimulation vs. null stimulation. KEY RESULTS: tVNS significantly increased activity in the insula, anterior cingulate, inferior and superior frontal gyri, caudate and putamen, and reduced activity in the hippocampi, occipital fusiform gyri, temporal pole, and middle temporal gyri, when compared to null stimulation (all corrected p < 0.005). tVNS increased activity in the anterior cingulate gyrus, left thalamus, caudate, and paracingulate gyrus and reduced activity in right thalamus, posterior cingulate cortex, and temporal fusiform cortex, when compared to sham stimulation (all corrected p < 0.005). Sham stimulation significantly increased activity in the insula and reduced activity in the posterior cingulate and paracingulate gyrus (all corrected p < 0.001), when contrasted to null stimulation. CONCLUSIONS: Brain effects of tVNS localize to regions associated with both physiological autonomic regulation and regions whose activity is modulated across numerous FGIDs, which may provide a neural basis for efficacy of this treatment. Functional activity differences between sham and null stimulation illustrate the importance of robust control procedures for future trials.

Representational ethical model calibration.

Equity is widely held to be fundamental to the ethics of healthcare. In the context of clinical decision-making, it rests on the comparative fidelity of the intelligence - evidence-based or intuitive - guiding the management of each individual patient. Though brought to recent attention by the individuating power of contemporary machine learning, such epistemic equity arises in the context of any decision guidance, whether traditional or innovative. Yet no general framework for its quantification, let alone assurance, currently exists. Here we formulate epistemic equity in terms of model fidelity evaluated over learnt multidimensional representations of identity crafted to maximise the captured diversity of the population, introducing a comprehensive framework for Representational Ethical Model Calibration. We demonstrate the use of the framework on large-scale multimodal data from UK Biobank to derive diverse representations of the population, quantify model performance, and institute responsive remediation. We offer our approach as a principled solution to quantifying and assuring epistemic equity in healthcare, with applications across the research, clinical, and regulatory domains.