Prompt-driven vision foundation models, such as the Segment Anything Model, have shown adaptability in computer vision, but their use in medical imaging remains challenging due to heterogeneous anatomy, artefacts, and low-contrast tumour boundaries. This is particularly difficult in whole-body tumour analysis, where models must transfer across modalities, anatomies, and tumour appearances. Here, we present Segment Any Tumour 3D (SAT3D), a lightweight volumetric foundation model for generalisable tumour segmentation across diverse medical imaging modalities, organs, and cohorts. SAT3D integrates a shifted-window vision transformer with critic-guided uncertainty-aware training, using confidence maps as dense prompts to guide boundary prediction in ambiguous regions. We benchmark SAT3D against vision foundation models, prompt-driven and task-specific methods across 11 public datasets. Trained on 17,075 three-dimensional volume-mask pairs, SAT3D shows robust generalisation, including in out-of-distribution settings, and is supported by a 3D-Slicer plugin for interactive segmentation, underscoring SAT3D’s potential as a scalable foundation model for medical image analysis. The application of prompt-driven vision foundation models to medical imaging, particularly whole-body tumour analysis, remains challenging. Here, the authors develop Segment Any Tumour 3D (SAT3D), a lightweight volumetric foundation model for robust and generalisable tumour segmentation across diverse medical imaging modalities, organs, and cohorts.

Variability ensures that complex biological systems, including the brain, are capable of responding to changing environmental demands. While the importance of neural variability in electrophysiological and hemodynamic aspects of brain activity is beginning to be understood, little is known about how variability in molecular activity influences brain function. Here, we examine how temporal variability in glucodynamics, or time-varying glucose use, is related to cognition in 35 younger and 43 older adults. Baseline metabolic rates of glucose indexed across the scan were not directly associated with cognition. Rather, higher glucodynamic variability, and its coherence into metabolic networks, was associated with better cognitive performance. Lower glucodynamic variability in aging was associated with altered metabolic network efficiency and reduced cognitive performance. Our results demonstrate for the first time that variability in cerebral glucose metabolism is biologically and functionally relevant to cognition and the network architecture of the brain. Cognition is influenced by time-varying glucose metabolism and its coherent fluctuations in metabolic networks. A loss of glucodynamics in aging reduces the efficiency of the metabolic connectome and contributes to reduced cognitive performance. The study of glucodynamics significantly advances our understanding of metabolic brain changes in health, aging, and disease.

The functional architecture of the brain is organised along continuous, macro-scale gradients. However, it is unknown if the metabolic architecture of the brain displays similar gradient characteristics. Here, we use functional positron emission tomography (fPET) with 18F-fluorodeoxyglucose (FDG) to characterise the metabolic connectivity gradients of the brain and determine how neurobiological mechanisms shape these gradients to support cognition across the adult lifespan. We identified four principal metabolic connectivity gradients, with the primary axis recapitulating the canonical unimodal-to-transmodal hierarchy from other imaging modalities. Subsequent gradients delineated specialised dimensions of metabolic organisation, including association system differentiation, hemispheric asymmetry, and sensory system segregation. These gradients were coupled to cortical thickness, baseline rates of glucose metabolism, blood flow, and gene expression related to energy metabolism, such that transmodal, control and default mode poles were more metabolically active, more interconnected, had greater cortical thickness, and were more strongly related to the expression of genes related to cellular energy production, than unimodal and sensory poles. A reduction in gradient strength at the gradient poles was associated with older age and predicted worse cognitive performance. We conclude that the metabolic organisation of the brain constitutes a genetically grounded, structurally and energetically constrained gradient hierarchy that supports cognitive function and undergoes a reorganisation in ageing. ### Competing Interest Statement The authors have declared no competing interest. Australian Research Council (ARC), DP25010302, FT250100206

Motion artifacts in magnetic resonance imaging (MRI) are one of the frequently occurring artifacts due to patient movements during scanning. Motion is estimated to be present in approximately 30% of clinical MRI scans; however, motion has not been explicitly modeled within deep learning image reconstruction models. Deep learning (DL) algorithms have been demonstrated to be effective for both the image reconstruction task and the motion correction task, but the two tasks are considered separately. The image reconstruction task involves removing undersampling artifacts such as noise and aliasing artifacts, whereas motion correction involves removing artifacts including blurring, ghosting, and ringing. In this work, we propose a novel method to simultaneously accelerate imaging and correct motion. This is achieved by integrating a motion module into the DL-based MRI reconstruction process, enabling detection and correction of motion. We model motion as a tightly integrated auxiliary layer in the DL model during training, making the DL model "motion-informed". During inference, image reconstruction is performed from undersampled raw k-space data using a trained motion-informed DL model. Experimental results demonstrate that the proposed motion-informed DL image reconstruction network outperformed the conventional image reconstruction network for motion-degraded MRI datasets.

The coupling between cerebral blood flow (CBF) and glucose metabolism (CMRGLC) is critical for maintaining brain function. However, sex differences in this relationship remain poorly understood, despite the heightened risk of cognitive decline from metabolic and vascular alterations in older women. Here, we address this gap by examining CBF-CMRGLC associations in 79 younger and older females and males using simultaneous MR/PET imaging and cognitive testing. Older adults exhibited weakened correlations between CBF and CMRGLC across functional networks. Sex moderated this decline, with older females showing significant negative CBF-CMRGLC associations, a pattern absent in older males and younger females. Individuals with stronger CBF-CMRGLC coupling performed better cognitively. Functional network parcellations (versus anatomical) better captured these sex- and age-specific effects. Our results support the idea that brain function depends not only on absolute metabolic substrate availability but on their coordinated use across functional networks. We conclude that the reduced cognitive performance of older adults is attributable to a loss of synchronized vascular and metabolic dynamics in functional networks. Other factors moderate this association, including sex and cardiometabolic health. Across older females, there are strong, negative network CBF-CMRGLC correlations, possibly reflecting a compensatory response in the face of attenuated rates of blood flow and glucose metabolism. The coupling of CBF and CMRGLC may serve as a biomarker for brain health and neurological conditions.

This study evaluated the hypothesis that 68Ga-PSMA-11 PET SUV, obtained via an advanced DL approach, correlates better with MR ADC maps than values from conventional PET-MR. Additionally, we aimed to identify the optimal SUV threshold for maximum correlation with ADC values. A cohort of 32 prostate cancer patients underwent CT and corresponding PET-MR imaging. The dataset underwent K-fold cross-validation, dividing it into four folds. In each fold, 24 patients were used for training, and 8 for validation to create DL models. ADC maps from 27 out of 32 patients were successfully aligned with T2 images for detailed analysis, revealing an inverse correlation (ρ = −0.20 to −0.51) between ADC and SUV values in prostate cancer zones. Statistically significant differences in mean SUV values were observed between PETMRI and PETDL. DL-based SUV values show a stronger correlation with ADC than conventional PET-MR values in our investigation.

Deep learning (DL) models are capable of successfully exploiting latent representations in MR data and have become state-of-the-art for accelerated MRI reconstruction. However, undersampling the measurements in k-space as well as the over- or under-parameterized and non-transparent nature of DL make these models exposed to uncertainty. Consequently, uncertainty estimation has become a major issue in DL MRI reconstruction. To estimate uncertainty, Monte Carlo (MC) inference techniques have become a common practice where multiple reconstructions are utilized to compute the variance in reconstruction as a measurement of uncertainty. However, these methods demand high computational costs as they require multiple inferences through the DL model. To this end, we introduce a method to estimate uncertainty during MRI reconstruction using a pixel classification framework. The proposed method, PixCUE (stands for Pixel Classification Uncertainty Estimation) produces the reconstructed image along with an uncertainty map during a single forward pass through the DL model. We demonstrate that this approach generates uncertainty maps that highly correlate with the reconstruction errors with respect to various MR imaging sequences and under numerous adversarial conditions. We also show that the estimated uncertainties are correlated to that of the conventional MC method. We further provide an empirical relationship between the uncertainty estimations using PixCUE and well-established reconstruction metrics such as NMSE, PSNR, and SSIM. We conclude that PixCUE is capable of reliably estimating the uncertainty in MRI reconstruction with a minimum additional computational cost.

In this study, we investigate jointly learning Hyperbolic and Euclidean space representations and match the consistency for semi-supervised medical image segmentation. We argue that for complex medical volumetric data, hyperbolic spaces are beneficial to model data inductive biases. We propose an approach incorporating the two geometries to co-train a variational encoder-decoder model with a Hyperbolic probabilistic latent space and a separate variational encoder-decoder model with a Euclidean probabilistic latent space with complementary representations, thereby bridging the gap of co-training across manifolds (Co-Manifold learning) in a principled manner. To capture complementary information and hierarchical relationships, we propose a Latent Space Loss aimed at maximizing disagreement between embeddings across manifolds. Additionally, we employ adversarial learning to enhance segmentation performance by guiding the network in hyperbolic latent space using confident regions identified by the network in Euclidean space. Conversely, the network in Euclidean space is informed by hyperbolic uncertainty, creating a dual uncertainty-aware framework that enables the two spaces to collaboratively learn confident regions from each other. Our proposed method achieves competitive results on two benchmarks for semi-supervised medical image segmentation on medical scans. The code is publicly available at: https://github.com/himashi92/Co-Manifold.

Computational competitions are the standard for benchmarking medical image analysis algorithms, but they typically use small curated test datasets acquired at a few centers, leaving a gap to the reality of diverse multicentric patient data. To this end, the Federated Tumor Segmentation (FeTS) Challenge represents the paradigm for real-world algorithmic performance evaluation. The FeTS challenge is a competition to benchmark (i) federated learning aggregation algorithms and (ii) state-of-the-art segmentation algorithms, across multiple international sites. Weight aggregation and client selection techniques were compared using a multicentric brain tumor dataset in realistic federated learning simulations, yielding benefits for adaptive weight aggregation, and efficiency gains through client sampling. Quantitative performance evaluation of state-of-the-art segmentation algorithms on data distributed internationally across 32 institutions yielded good generalization on average, albeit the worst-case performance revealed data-specific modes of failure. Similar multi-site setups can help validate the real-world utility of healthcare AI algorithms in the future.

Implicit Neural Representations (INRs) have recently advanced the field of deep learning due to their ability to learn continuous representations of signals without the need for large training datasets. Although INR methods have been studied for medical image super-resolution, their adaptability to localized priors in medical images has not been extensively explored. Medical images contain rich anatomical divisions that could provide valuable local prior information to enhance the accuracy and robustness of INRs. In this work, we propose a novel framework, referred to as the Semantically Conditioned INR (SeCo-INR), that conditions an INR using local priors from a medical image, enabling accurate model fitting and interpolation capabilities to achieve super-resolution. Our framework learns a continuous representation of the semantic segmentation features of a medical image and utilizes it to derive the optimal INR for each semantic region of the image. We tested our framework using several medical imaging modalities and achieved higher quantitative scores and more realistic super-resolution outputs compared to state-of-the-art methods.

This study compares volumetric measurements of various brain regions using different magnetic resonance imaging (MRI) modalities and deep learning models, specifically 3T MRI, ultra-low field (ULF) MRI at 64mT, and AI-enhanced ULF MRI using SynthSR and HiLoResGAN. The aim is to evaluate the alignment and agreement among field strengths and ULF MRI with and without AI. Descriptive statistics, paired t-tests, effect size analyses, and regression analyses are employed to assess the relationships and differences between modalities. The results indicate that volumetric measurements derived from 64mT MRI deviate significantly from those obtained using 3T MRI. By leveraging SynthSR and LoHiResGAN models, these deviations are reduced, bringing the volumetric estimates closer to those obtained from 3T MRI, which serves as the reference standard for brain volume quantification. These findings highlight that deep learning models can reduce systematic differences in brain volume measurements across field strengths, providing potential solutions to minimize bias in imaging studies.

Psychedelics can profoundly alter consciousness by reorganising brain connectivity; however, their effects are context-sensitive. To understand how this reorganisation depends on the context, we collected and comprehensively analysed the largest psychedelic neuroimaging dataset to date. Sixty-two adults were scanned with fMRI and EEG during rest and naturalistic stimuli (meditation, music, and visual), before and after ingesting 19 mg of psilocybin. Half of the participants ranked the experience among the five most meaningful of their lives. Under psilocybin, fMRI and EEG signals recorded during eyes-closed conditions became similar to those recorded during an eyes-open condition. This change manifested as an increase in global functional connectivity in associative regions and a decrease in sensory areas. For the first time, we used machine learning to directly link the subjective effects of psychedelics to neural activity patterns characterised by low-dimensional embeddings. We show that psilocybin reorganised these low-dimensional trajectories into cohesive patterns of brain activity that were structured by context and quality of subjective experience, with stronger self- and boundary-related effects - which were linked to day-after mindset changes - leading to more structured and distinct neural representations. This reorganisation induces a state of 'embeddedness' - a coherent integration of brain networks that normally segregate internal and external processing - dissolving perceptual boundaries in a way that aligns neural dynamics with context. Beyond its transient expression, embeddedness serves as a construct for understanding the subjective and therapeutic effects of psychedelics. These findings provide a new account of the large-scale neurocognitive effects of psychedelics and demonstrate the utility of using machine learning methods in assessing state- and context-dependent neural dynamics and their association with psychological outcomes. ### Competing Interest Statement The authors have declared no competing interest.

BACKGROUND:Retinal vessel calibres (RVCs) are non-invasive markers of microvascular health and may serve as accessible indicators of cerebral small vessel disease (CSVD) and future cognitive impairment. This study examines whether RVCs are associated with cognitive decline, and how these associations compare with those observed for white matter hyperintensities (WMH), a known marker of CSVD. METHODS:Data were analysed from community-dwelling participants aged 70+ in the ASPREE trial and sub-studies, free of dementia and cardiovascular disease at baseline. RVCs were measured from fundus photography and WMH volumes from 3 T magnetic resonance imaging. Covariate-adjusted linear mixed-effects models assessed cognitive trajectories relative to baseline RVCs and WMH volumes. Cross-sectional associations between baseline RVCs and WMHs were examined via linear regression. RESULTS:This study included 3540 participants with RVC data and 489 with WMH data (median [IQR] age: 73.2 [71.4-76.3] and 72.5 [71.2-75.4] years; female: 52.9% and 47.6%) over a median follow-up of 7.4 [IQR 5.5-8.5] and 3.8 [IQR 2.9-5.3] years, respectively. Baseline RVCs were not significantly associated with cognitive trajectories nor with baseline WMHs. Larger baseline WMH volumes were associated with greater global (Modified Mini-Mental State Examination) decline (mean 0.40 points/year; 95% CI 0.57, 0.22) and declines in delayed memory (HVLT-r) (-0.13 [-0.22, -0.04]), psychomotor function (Symbol Digit Modalities Test) (-0.29 [-0.52, -0.07]) and to a lesser extent, executive function (Controlled Oral Word Association Test) (-0.09 [95% CI -0.22, 0.03]). CONCLUSION:In contrast to WMH volumes, RVCs were not associated with cognitive decline. Exploring longitudinal changes in a broader range of retinal and brain biomarkers may provide deeper insights into the relationship between ocular and cerebral biomarkers in CSVD and clinical outcomes.

Recently developed high temporal resolution functional (18F)-fluorodeoxyglucose positron emission tomography (fPET) offers promise as a method for indexing the dynamic metabolic state of the brain in vivo by directly measuring a time series of metabolism at the post-synaptic neuron. This is distinct from functional magnetic resonance imaging (fMRI) that reflects a combination of metabolic, haemodynamic and vascular components of neuronal activity. The value of using fPET to understand healthy brain ageing and cognition over fMRI is currently unclear. Here, we use simultaneous fPET/fMRI to compare metabolic and functional connectivity and test their predictive ability for ageing and cognition. Whole-brain fPET connectomes showed moderate topological similarities to fMRI connectomes in a cross-sectional comparison of 40 younger (mean age 27.9 years; range 20-42) and 46 older (mean 75.8; 60-89) adults. There were more age-related within- and between-network connectivity and graph metric differences in fPET than fMRI. fPET was also associated with performance in more cognitive domains than fMRI. These results suggest that ageing is associated with a reconfiguration of metabolic connectivity that differs from haemodynamic alterations. We conclude that metabolic connectivity has greater predictive utility for age and cognition than functional connectivity and that measuring glucodynamic changes has promise as a biomarker for age-related cognitive decline. Deery et al. report that there is a reconfiguration of metabolic brain networks in ageing that is different to haemodynamic alterations and is more strongly linked to cognitive performance. They conclude that metabolic connectivity offers utility to understand brain network changes in health and disease.

In Magnetic Resonance Imaging (MRI), image acquisitions are often undersampled in the measurement domain to accelerate the scanning process, at the expense of image quality. However, image quality is a crucial factor that influences the accuracy of clinical diagnosis; hence, high-quality image reconstruction from undersampled measurements has been a key area of research. Recently, deep learning (DL) methods have emerged as the state-of-the-art for MRI reconstruction, typically involving deep neural networks to transform undersampled MRI images into high-quality MRI images through data-driven processes. Nevertheless, there is clear and significant room for improvement in undersampled DL MRI reconstruction to meet the high standards required for clinical diagnosis, in terms of eliminating aliasing artifacts and reducing image noise. In this paper, we introduce a self-supervised pretraining procedure using contrastive learning to improve the accuracy of undersampled DL MRI reconstruction. We use contrastive learning to transform the MRI image representations into a latent space that maximizes mutual information among different undersampled representations and optimizes the information content at the input of the downstream DL reconstruction models. Our experiments demonstrate improved reconstruction accuracy across a range of acceleration factors and datasets, both quantitatively and qualitatively. Furthermore, our extended experiments validate the proposed framework's robustness under adversarial conditions, such as measurement noise, different k-space sampling patterns, and pathological abnormalities, and also prove the transfer learning capabilities on MRI datasets with completely different anatomy. Additionally, we conducted experiments to visualize and analyze the properties of the proposed MRI contrastive learning latent space.

People with insulin resistance are at increased risk for cognitive decline. Insulin resistance has previously been considered primarily a condition of ageing but it is increasingly seen in younger adults. It is possible that impaired insulin function in early adulthood has both proximal effects and moderates or even accelerates changes in cerebral metabolism in ageing. Thirty-six younger (mean 27.8 years) and 43 older (mean 75.5) participants completed a battery of tests, including blood sampling, cognitive assessment and a simultaneous PET/MR scan. Cortical thickness and cerebral metabolic rates of glucose were derived for 100 regions and 17 functional networks. Older adults had lower rates of regional cerebral glucose metabolism than younger adults across the brain even after adjusting for lower cortical thickness in older adults. Higher fasting blood glucose was also associated with lower regional cerebral glucose metabolism in older adults. In younger adults, higher insulin resistance was associated with lower rates of regional cerebral glucose metabolism but this was not seen in older adults. The largest effects of insulin resistance in younger adults were in prefrontal, parietal and temporal regions; and in the control, salience ventral attention, default and somatomotor networks. Higher rates of network glucose metabolism were associated with lower reaction time and psychomotor speed. Higher levels of insulin resistance were associated with lower working memory. Our results underscore the importance of insulin sensitivity and glycaemic control to brain health and cognitive function across the adult lifespan, even in early adulthood.

As intensities of MRI volumes are inconsistent across institutes, it is essential to extract universal features of multi-modal MRIs to precisely segment brain tumors. In this concept, we propose a volumetric vision transformer that follows two windowing strategies in attention for extracting fine features and local distributional smoothness (LDS) during model training inspired by virtual adversarial training (VAT) to make the model robust. We trained and evaluated network architecture on the FeTS Challenge 2022 dataset. Our performance on the online evaluation is as follows: Dice Similarity Score of 85.70%, 90.59% and 87.27%; Hausdorff Distance (95%) of 10.46 mm, 7.40 mm, 12.66mm for the enhancing tumor, whole tumor, and tumor core, respectively. Overall, the experimental results verify our method's effectiveness by yielding better performance in segmentation accuracy for each tumor sub-region. Our code implementation is publicly available.

BackgroundThe dentate nuclei of the cerebellum are key sites of neuropathology in Friedreich ataxia (FRDA). Reduced dentate nucleus volume and increased mean magnetic susceptibility, a proxy of iron concentration, have been reported by magnetic resonance imaging studies in people with FRDA. Here, we investigate whether these changes are regionally heterogeneous.MethodsQuantitative susceptibility mapping data were acquired from 49 people with FRDA and 46 healthy controls. The dentate nuclei were manually segmented and analyzed using three dimensional vertex-based shape modeling and voxel-based assessments to identify regional changes in morphometry and susceptibility, respectively.ResultsIndividuals with FRDA, relative to healthy controls, showed significant bilateral surface contraction most strongly at the rostral and caudal boundaries of the dentate nuclei. The magnitude of this surface contraction correlated with disease duration, and to a lesser extent, ataxia severity. Significantly greater susceptibility was also evident in the FRDA cohort relative to controls, but was instead localized to bilateral dorsomedial areas, and also correlated with disease duration and ataxia severity.ConclusionsChanges in the structure of the dentate nuclei in FRDA are not spatially uniform. Atrophy is greatest in areas with high gray matter density, whereas increases in susceptibility-reflecting iron concentration, demyelination, and/or gliosis-predominate in the medial white matter. These findings converge with established histological reports and indicate that regional measures of dentate nucleus substructure are more sensitive measures of disease expression than full-structure averages. Biomarker development and therapeutic strategies that directly target the dentate nuclei, such as gene therapies, may be optimized by targeting these areas of maximal pathology. (c) 2024 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Chronic liver disease is responsible for significant morbidity and mortality worldwide. Abdominal computed tomography (CT) and magnetic resonance imaging (MRI) can fully visualise the liver and adjacent structures in the upper abdomen providing a reproducible assessment of the liver and biliary system and can detect features of portal hypertension. Subjective interpretation of CT and MRI in the assessment of liver parenchyma for early and advanced stages of fibrosis (pre-cirrhosis), as well as severity of portal hypertension, is limited. Quantitative and reproducible measurements of hepatic and splenic volumes have been shown to correlate with fibrosis staging, clinical outcomes, and mortality. In this review, we will explore the role of volumetric measurements in relation to diagnosis, assessment of severity and prediction of outcomes in chronic liver disease patients. We conclude that volumetric analysis of the liver and spleen can provide important information in such patients, has the potential to stratify patients' stage of hepatic fibrosis and disease severity, and can provide critical prognostic information. Critical relevance statement This review highlights the role of volumetric measurements of the liver and spleen using CT and MRI in relation to diagnosis, assessment of severity, and prediction of outcomes in chronic liver disease patients. Key Points Volumetry of the liver and spleen using CT and MRI correlates with hepatic fibrosis stages and cirrhosis. Volumetric measurements correlate with chronic liver disease outcomes. Fully automated methods for volumetry are required for implementation into routine clinical practice.

Background: The neurological phenotype of Friedreich ataxia (FRDA) is characterized by neurodegeneration and neuroinflammation in the cerebellum and brainstem. Novel neuroimaging approaches quantifying brain free-water using diffusion magnetic resonance imaging (dMRI) are potentially more sensitive to these processes than standard imaging markers.Objectives: To quantify the extent of free-water and microstructural change in FRDA-relevant brain regions using neurite orientation dispersion and density imaging (NODDI), and bitensor diffusion tensor imaging (btDTI).Method: Multi-shell dMRI was acquired from 14 individuals with FRDA and 14 controls. Free-water measures from NODDI (FISO) and btDTI (FW) were compared between groups in the cerebellar cortex, dentate nuclei, cerebellar peduncles, and brainstem. The relative sensitivity of the free-water measures to group differences was compared to microstructural measures of NODDI intracellular volume, free-water corrected fractional anisotropy, and conventional uncorrected fractional anisotropy.Results: In individuals with FRDA, FW was elevated in the cerebellar cortex, peduncles (excluding middle), dentate, and brainstem (P < 0.005). FISO was elevated primarily in the cerebellar lobules (P < 0.001). On average, FW effect sizes were larger than all other markers (mean eta(rho) (2) = 0.43), although microstructural measures also had very large effects in the superior and inferior cerebellar peduncles and brainstem (eta(rho) (2) > 0.37). Across all regions and metrics, effect sizes were largest in the superior cerebellar peduncles (eta(rho) (2) > 0.46).Conclusions: Multi-compartment diffusion measures of free-water and neurite integrity distinguish FRDA from controls with large effects. Free-water magnitude in the brainstem and cerebellum provided the greatest distinction between groups. This study supports further applications of multi-compartment diffusion modeling, and investigations of free-water as a measure of disease expression and progression in FRDA. (c) 2023 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.