Multimodal learning has attracted much attention in recent years due to its ability to effectively utilize data features from a variety of different modalities. Diagnosing the vulnerability of atherosclerotic plaques directly from carotid 3D MRI images is relatively challenging for both radiologists and conventional 3D vision networks. In clinical practice, radiologists assess patient conditions using a multimodal approach that incorporates various imaging modalities and domain-specific expertise, paving the way for the creation of multimodal diagnostic networks. In this paper, we have developed an effective strategy to leverage radiologists' domain knowledge to automate the diagnosis of carotid plaque vulnerability through Variation inference and Multimodal knowledge Distillation (VMD). This method excels in harnessing cross-modality prior knowledge from limited image annotations and radiology reports within training data, thereby enhancing the diagnostic network's accuracy for unannotated 3D MRI images. We conducted in-depth experiments on the dataset collected in-house and verified the effectiveness of the VMD strategy we proposed.

Time-series forecasting and causal discovery are central in neuroscience, as predicting brain activity and identifying causal relationships between neural populations and circuits can shed light on the mechanisms underlying cognition and disease. With the rise of foundation models, an open question is how they compare to traditional methods for brain signal forecasting and causality analysis, and whether they can be applied in a zero-shot setting. In this work, we evaluate a foundation model against classical methods for inferring directional interactions from spontaneous brain activity measured with functional magnetic resonance imaging (fMRI) in humans. Traditional approaches often rely on Wiener-Granger causality. We tested the forecasting ability of the foundation model in both zero-shot and fine-tuned settings, and assessed causality by comparing Granger-like estimates from the model with standard Granger causality. We validated the approach using synthetic time series generated from ground-truth causal models, including logistic map coupling and Ornstein-Uhlenbeck processes. The foundation model achieved competitive zero-shot forecasting fMRI time series (mean absolute percentage error of 0.55 in controls and 0.27 in patients). Although standard Granger causality did not show clear quantitative differences between models, the foundation model provided a more precise detection of causal interactions. Overall, these findings suggest that foundation models offer versatility, strong zero-shot performance, and potential utility for forecasting and causal discovery in time-series data.

Time-series forecasting and causal discovery are central in neuroscience, as predicting brain activity and identifying causal relationships between neural populations and circuits can shed light on the mechanisms underlying cognition and disease. With the rise of foundation models, an open question is how they compare to traditional methods for brain signal forecasting and causality analysis, and whether they can be applied in a zero-shot setting. In this work, we evaluate a foundation model against classical methods for inferring directional interactions from spontaneous brain activity measured with functional magnetic resonance imaging (fMRI) in humans. Traditional approaches often rely on Wiener-Granger causality. We tested the forecasting ability of the foundation model in both zero-shot and fine-tuned settings, and assessed causality by comparing Granger-like estimates from the model with standard Granger causality. We validated the approach using synthetic time series generated from ground-truth causal models, including logistic map coupling and Ornstein-Uhlenbeck processes. The foundation model achieved competitive zero-shot forecasting fMRI time series (mean absolute percentage error of 0.55 in controls and 0.27 in patients). Although standard Granger causality did not show clear quantitative differences between models, the foundation model provided a more precise detection of causal interactions. Overall, these findings suggest that foundation models offer versatility, strong zero-shot performance, and potential utility for forecasting and causal discovery in time-series data.

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that currently affects over 55 million individuals worldwide. Conventional diagnostic approaches often rely on subjective clinical assessments and isolated biomarkers, limiting their accuracy and early-stage effectiveness. With the rising global burden of AD, there is an urgent need for objective, automated tools that enhance diagnostic precision using neuroimaging data. This study proposes a novel diagnostic framework combining a fine-tuned VGG19 deep convolutional neural network with an eXtreme Gradient Boosting (XGBoost) classifier. The model was trained and validated on the OASIS MRI dataset (Dataset 2), which was manually balanced to ensure equitable class representation across the four AD stages. The VGG19 model was pre-trained on ImageNet and fine-tuned by unfreezing its last ten layers. Data augmentation strategies, including random rotation and zoom, were applied to improve generalization. Extracted features were classified using XGBoost, incorporating class weighting, early stopping, and adaptive learning. Model performance was evaluated using accuracy, precision, recall, F1-score, and ROC-AUC. The proposed VGG19-XGBoost model achieved a test accuracy of 99.6%, with an average precision of 1.00, a recall of 0.99, and an F1-score of 0.99 on the balanced OASIS dataset. ROC curves indicated high separability across AD stages, confirming strong discriminatory power and robustness in classification. The integration of deep feature extraction with ensemble learning demonstrated substantial improvement over conventional single-model approaches. The hybrid model effectively mitigated issues of class imbalance and overfitting, offering stable performance across all dementia stages. These findings suggest the method's practical viability for clinical decision support in early AD diagnosis. This study presents a high-performing, automated diagnostic tool for Alzheimer's disease based on neuroimaging. The VGG19-XGBoost hybrid architecture demonstrates exceptional accuracy and robustness, underscoring its potential for real-world applications. Future work will focus on integrating multimodal data and validating the model on larger and more diverse populations to enhance clinical utility and generalizability.

BackgroundBrain atrophy is a hallmark of multiple sclerosis (MS). For clinical translatability and individual-level predictions, brain atrophy needs to be put into context of the broader population, using reference or normative models. MethodsReference models of MRI-derived brain volumes were established from a large healthy control (HC) multi-cohort dataset (N=63 115, 51% females). The reference models were applied to two independent MS cohorts (N=362, T1w-scans=953, follow-up time up to 12 years) to assess deviations from the reference, defined as Z-values. We assessed the overlap of deviation profiles and their stability over time using individual-level transitions towards or out of significant reference deviation states (|Z|>1{middle dot}96). A negative binomial model was used for case-control comparisons of the number of extreme deviations. Linear models were used to assess differences in Z-score deviations between MS and propensity-matched HCs, and associations with clinical scores at baseline and over time. The utilized normative BrainReference models, scripts and usage instructions are freely available. FindingsWe identified a temporally stable, brain morphometric phenotype of MS. The right and left thalami most consistently showed significantly lower-than-reference volumes in MS (25% and 26% overlap across the sample). The number of such extreme smaller-than-reference values was 2{middle dot}70 in MS compared to HC (4{middle dot}51 versus 1{middle dot}67). Additional deviations indicated stronger disability (Expanded Disability Status Scale: {beta}=0{middle dot}22, 95% CI 0{middle dot}12 to 0{middle dot}32), Paced Auditory Serial Addition Test score ({beta}=-0{middle dot}27, 95% CI -0{middle dot}52 to -0{middle dot}02), and Fatigue Severity Score ({beta}=0{middle dot}29, 95% CI 0{middle dot}05 to 0{middle dot}53) at baseline, and over time with EDSS ({beta}=0{middle dot}07, 95% CI 0{middle dot}02 to 0{middle dot}13). We additionally provide detailed maps of reference-deviations and their associations with clinical assessments. InterpretationWe present a heterogenous brain phenotype of MS which is associated with clinical manifestations, and particularly implicating the thalamus. The findings offer potential to aid diagnosis and prognosis of MS. FundingNorwegian MS-union, Research Council of Norway (#223273; #324252); the South-Eastern Norway Regional Health Authority (#2022080); and the European Unions Horizon2020 Research and Innovation Programme (#847776, #802998). Research in contextO_ST_ABSEvidence before this studyC_ST_ABSReference values and normative models have yet to be widely applied to neuroimaging assessments of neurological disorders such as multiple sclerosis (MS). We conducted a literature search in PubMed and Embase (Jan 1, 2000-September 12, 2025) using the terms "MRI" AND "multiple sclerosis", with and without the keywords "normative model*" and "atrophy", without language restrictions. While normative models have been applied in psychiatric and developmental disorders, few studies have addressed their use in neurological conditions. Existing MS research has largely focused on global atrophy and has not provided regional reference charts or established links to clinical and cognitive outcomes. Added value of this studyWe provide regionally detailed brain morphometry maps derived from a heterogeneous MS cohort spanning wide ranges of age, sex, clinical phenotype, disease duration, disability, and scanner characteristics. By leveraging normative modelling, our approach enables individualised brain phenotyping of MS in relation to a population based normative sample. The analyses reveal clinically meaningful and spatially consistent patterns of smaller brain volumes, particularly in the thalamus and frontal cortical regions, which are linked to disability, cognitive impairment, and fatigue. Robustness across scanners, centres, and longitudinal follow-up supports the stability and generalisability of these findings to real-world MS populations. Implications of all the available evidenceNormative modelling offers an individualised, sensitive, and interpretable approach to quantifying brain structure in MS by providing individual-specific reference values, supporting earlier detection of neurodegeneration and improved patient stratification. A consistent pattern of thalamic and fronto-parietal deviations defines a distinct morphometric profile of MS, with potential utility for early and personalised diagnosis and disease monitoring in clinical practice and clinical trials.

Idiopathic normal pressure hydrocephalus (iNPH) is a common and debilitating condition whose diagnosis is made challenging due to the unspecific and common clinical presentation. The aim of our study was to determine if data driven patterns of cerebrospinal fluid (CSF) distribution can be used to predict iNPH diagnosis and response to treatment. We established a cohort of iNPH patients and age/sex-matched controls. We used Non-negative Matrix Factorization (NMF) on CSF probability maps from segmentation of T1-weighted MRI to obtain patterns or components of CSF distribution across participants and a load on each component in each participant. Visual assessment of morphologic phenotype was performed by a neuroradiologist, and clinical symptom improvement was assessed via retrospective chart review. We used the NMF component loads to predict diagnosis and clinical outcome after ventriculoperitoneal shunt placement for treatment of iNPH. Similar models were developed using manual Evan's index and callosal angle measurements. We included 98 iNPH patients and 98 controls split into test (20 %) and train (80 %) sets. The optimal NMF decomposition identified 7 patterns of CSF distribution in our cohort. Accuracy for predicting a clinical diagnosis of iNPH using the automated NMF model was 96 %/97 % in the train/test sets, which was similar to the performance of the manual measure models (92 %/97 %). Visualizing the voxels that contributed most to the NMF models revealed that the voxels most associated with a disproportionately enlarged subarachnoid space hydrocephalus (DESH) were the ones with higher probability of iNPH diagnosis. Neither NMF nor manual metrics performed well for prediction of qualitative clinical outcomes. NMF-generated patterns of CSF distribution showed high accuracy in discerning individuals with iNPH from controls. The patterns most relying on DESH features showed highest potential for independently predicting NPH diagnosis. The algorithm we proposed should not be perceived as a replacement for human expertise but rather as an additional tool to assist clinicians in achieving accurate diagnoses.

Background and ObjectivesIdentifying MS in children early and distinguishing it from other neuroinflammatory conditions of childhood is critical, as early therapeutic intervention can improve outcomes. The anterior visual pathway has been demonstrated to be of central importance in diagnostic considerations for MS and has recently been identified as a fifth topography in the McDonald Diagnostic Criteria for MS. Optical coherence tomography (OCT) provides high-resolution retinal imaging and reflects the structural integrity of the retinal nerve fiber and ganglion cell inner plexiform layers. Whether multimodal deep learning models can use OCT alone to diagnose pediatric MS (POMS) is unknown. MethodsWe analyzed 3D OCT scans collected prospectively through the Neuroinflammatory Registry of the Hospital for Sick Children (REB#1000005356). Raw macular and optic nerve head images, and 52 automatically segmented features were included. We evaluated three classification approaches: (1) deep learning models (e.g. ResNet, DenseNet) for representation learning followed by classical ML classifiers, (2) ML models trained on OCT-derived features, and (3) multimodal models combining both via early and late fusion. ResultsScans from individuals with POMS (onset 16.0 {+/-} 3.1 years, 51.0%F; 211 scans) and 29 children with non-inflammatory neurological conditions (13.1 {+/-} 4.0 years, 69.0%F, 52 scans) were included. The early fusion model achieved the highest performance (AUC: 0.87, F1: 0.87, Accuracy: 90%), outperforming both unimodal and late fusion models. The best unimodal feature-based model (SVC) yielded an AUC of 0.84, F1 of 0.85 and an accuracy of 85%, while the best image-based model (ResNet101 with Random Forest) achieved an AUC of 0.87, F1 of 0.79, and accuracy of 84%. Late fusion underperformed, reaching 82% accuracy but failing in the minority class. DiscussionMultimodal learning with early fusion significantly enhances diagnostic performance by combining spatial retinal information with clinically relevant structural features. This approach captures complementary patterns associated with MS pathology and shows promise as an AI-driven tool to support pediatric neuroinflammatory diagnosis.

Congestive hepatopathy, also known as nutmeg liver, is liver damage secondary to chronic heart failure (HF). Its morphological characteristics in terms of medical imaging are not defined and remain unclear. To leverage machine learning to capture imaging features of congestive hepatopathy using incidentally acquired computed tomography (CT) scans. We retrospectively analyzed 179 chronic HF patients who underwent echocardiography and CT within one year. Right HF severity was classified into three grades. Liver CT images at the paraumbilical vein level were used to develop a ResNet-based machine learning model to predict tricuspid regurgitation (TR) severity. Model accuracy was compared with that of six gastroenterology and four radiology experts. In the included patients, 120 were male (mean age: 73.1 ± 14.4 years). The accuracy of the results predicting TR severity from a single CT image for the machine learning model was significantly higher than the average accuracy of the experts. The model was found to be exceptionally reliable for predicting severe TR. Deep learning models, particularly those using ResNet architectures, can help identify morphological changes associated with TR severity, aiding in early liver dysfunction detection in patients with HF, thereby improving outcomes.

Identifying new disease-related patterns in medical imaging data with the help of machine learning enlarges the vocabulary of recognizable findings. This supports diagnostic and prognostic assessment. However, image appearance varies not only due to biological differences, but also due to imaging technology linked to vendors, scanning- or re- construction parameters. The resulting domain shifts impedes data representation learning strategies and the discovery of biologically meaningful cluster appearances. To address these challenges, we introduce an approach to actively learn the domain shift via post-hoc rotation of the data latent space, enabling disentanglement of biological and technical factors. Results on real-world heterogeneous clinical data showcase that the learned disentangled representation leads to stable clusters representing tissue-types across different acquisition settings. Cluster consistency is improved by +19.01% (ARI), +16.85% (NMI), and +12.39% (Dice) compared to the entangled representation, outperforming four state-of-the-art harmonization methods. When using the clusters to quantify tissue composition on idiopathic pulmonary fibrosis patients, the learned profiles enhance Cox survival prediction. This indicates that the proposed label-free framework facilitates biomarker discovery in multi-center routine imaging data. Code is available on GitHub https://github.com/cirmuw/latent-space-rotation-disentanglement.

In real world clinical environments, training and applying deep learning models on multi-modal medical imaging data often struggles with partially incomplete data. Standard approaches either discard missing samples, require imputation or repurpose dropout learning schemes, limiting robustness and generalizability. To address this, we propose a hypernetwork-based method that dynamically generates task-specific classification models conditioned on the set of available modalities. Instead of training a fixed model, a hypernetwork learns to predict the parameters of a task model adapted to available modalities, enabling training and inference on all samples, regardless of completeness. We compare this approach with (1) models trained only on complete data, (2) state of the art channel dropout methods, and (3) an imputation-based method, using artificially incomplete datasets to systematically analyze robustness to missing modalities. Results demonstrate superior adaptability of our method, outperforming state of the art approaches with an absolute increase in accuracy of up to 8% when trained on a dataset with 25% completeness (75% of training data with missing modalities). By enabling a single model to generalize across all modality configurations, our approach provides an efficient solution for real-world multi-modal medical data analysis.