model-to-data alignment

To address the challenge of establishing stable inter-subject local correspondences in anatomical shape statistics—traditionally reliant on explicit registration—we propose a registration-free, boundary- and interior-consistent geometric modeling framework. Our method models target deformations as diffeomorphic transformations of an ellipsoid, embedded within a globally optimized skeleton-driven fitting scheme that simultaneously constructs a consistent coordinate system on both the object’s boundary and interior. We further introduce an evolutionary s-rep representation, the first to encode intrinsic geometric features directly in the fitted coordinate space, enabling robust point-wise correspondence across subjects without mesh alignment. The approach integrates differential-geometric deformation modeling, skeleton-guided fitting, and boundary-driven intrinsic coordinate generation. In hippocampal disease classification, it significantly outperforms two state-of-the-art methods, demonstrating superior discriminative power and statistical stability of the learned features.

Current public brain MRI datasets lack systematic evaluation in scale, diversity, and consistency, hindering foundation model generalizability. To address this, we conduct the first cross-dataset quantitative profiling study across 54 publicly available datasets (>530,000 scans), establishing a multi-level assessment framework spanning dataset-, image-, and feature-level analyses. Our methodology integrates modality distribution statistics, voxel spacing and intensity quantification, preprocessing pipeline variability analysis, and validation within the 3D DenseNet121 feature space. Results reveal critical issues: dominance of healthy controls, uneven disease coverage, substantial geometric and intensity heterogeneity, and residual covariate shift post-preprocessing. We identify the need for perceptually informed preprocessing combined with domain-adaptive modeling. This work provides a reproducible, structured evaluation paradigm to guide data curation and algorithm design for brain MRI foundation models.

This work addresses the challenges in neuroimaging research caused by data heterogeneity and incompatible tool interfaces, which often necessitate redundant implementations of routine operations. To overcome these limitations, the authors propose a unified, open-source Python framework that, for the first time, integrates volumetric, cortical surface, and streamline data within a single system. Through an object-oriented design, it provides consistent interfaces for loading, processing, and exporting data, while supporting BIDS compliance, FreeSurfer integration, diffusion MRI analysis, and GPU-accelerated visualization. The framework includes built-in color map and lookup table management, brain parcellation, multi-surface rendering, and tractography capabilities, enabling code-free workflow customization via JSON configuration. Developed for Python 3.9–3.12, it supports standard formats such as NIfTI, GIFTI, and TCK/TRK, substantially lowering technical barriers. The code is publicly available with comprehensive documentation and examples.

This study addresses the lack of a statistically rigorous framework in current neuroimaging research for testing hypotheses about associations between structural and functional brain data. The authors propose the first explicit Bayesian hypothesis testing approach that integrates fMRI-derived functional brain networks with regional structural measurements through a hierarchical Bayesian model. This method explicitly models the relationship between structure and function while providing full posterior uncertainty quantification. It facilitates the integration of heterogeneous data types and incorporation of prior information, demonstrating robust and efficient detection of structure–function associations across varying signal-to-noise ratios, numbers of brain regions, and types of structural measures. The proposed approach substantially outperforms existing methods in both accuracy and reliability.

This study addresses the challenge of comparing high-dimensional neural representations across neuroscience and artificial intelligence: specifically, how to select similarity measures that best reveal functional correspondences and divergences. We systematically evaluate eight mainstream representational similarity metrics—including linear CKA, Procrustes distance, CCA, inner-product kernel, and nearest-neighbor alignment—against behavioral functional alignment (e.g., recognition accuracy, generalization, robustness) as a ground-truth benchmark. Our evaluation spans both biological neural data and artificial neural network models. Results show that geometry-sensitive metrics—particularly linear CKA and Procrustes distance—consistently outperform predictive metrics, achieving superior alignment with human behavioral performance and effectively distinguishing trained versus untrained models. In contrast, linear predictivity exhibits only moderate behavioral alignment. This work establishes the first behavior-driven representational similarity benchmark, providing a principled, cross-domain methodology for mechanistic interpretation and comparative analysis of neural computation.

This work addresses the limited generalization performance in cross-subject brain functional decoding caused by inter-individual variability in neural responses. To overcome this challenge, the authors propose SpectralOT, a novel method that, for the first time, integrates spectral features of the Laplace–Beltrami operator into functional data and leverages optimal transport theory to construct a geometry-aware whole-brain alignment framework. By explicitly incorporating cortical geometric structure during functional alignment, the approach enhances computational efficiency while preserving anatomical consistency. Experimental results demonstrate that SpectralOT significantly improves the generalization capability of cross-subject decoding models, offering a new paradigm for high-precision brain functional analysis.

This study addresses the limitation of relying solely on prediction accuracy to assess alignment between visual models and human brain responses, as such metrics often obscure which reproducible response dimensions are shared. To overcome this, the authors propose a unified evaluation framework that quantifies the degree to which models or cross-subject brain signals recover reproducible dimensions within a target neural response space, using target-space recovery profiles rather than scalar accuracy measures. Integrating repeated fMRI measurements, cross-run splits, and both brain–brain and model–brain predictive modeling, the approach reveals a low-dimensional, reproducible response structure in early-to-mid-level visual cortex during naturalistic viewing. Notably, despite similar prediction accuracies, pretrained and randomly initialized models exhibit markedly distinct recovery profiles, uncovering alignment differences masked by conventional accuracy metrics.

Existing learning-based cortical surface parcellation methods lack in-depth analysis of performance improvement mechanisms—particularly their interplay with registration and atlas propagation. Method: We propose the first end-to-end joint cortical registration and parcellation framework, featuring deep coupling between the two tasks: a learnable atlas propagation module and a shallow fine-tuning subnetwork; a lightweight geometric-feature-driven architecture (using sulcal depth and curvature) that jointly optimizes diffeomorphic registration and label propagation. Results: On the Mindboggle dataset, our method achieves Dice scores exceeding 90%, significantly outperforming both conventional and state-of-the-art learning-based approaches. Ablation studies confirm registration quality as the key bottleneck limiting parcellation accuracy. Our framework enhances anatomical consistency and label fidelity of cortical parcellations while improving statistical power of brain atlases—thereby providing robust support for clinical applications such as neurosurgical planning.

Neuroimaging meta-analyses frequently suffer from statistical unreliability due to small sample sizes, and conventional methods struggle to capture the intrinsic hierarchical organization of brain function and cross-modal semantic relationships. To address these limitations, we propose the first hyperbolic multi-level meta-analysis framework, jointly embedding brain activation maps and literature text into the Lorentz model of hyperbolic space. This enables semantic alignment, cross-modal hierarchical guidance, and preservation of the hierarchical structure of neural activation patterns. Crucially, our approach is the first to explicitly incorporate hyperbolic geometry into neuroimaging meta-analysis, leveraging its natural capacity to model tree-like functional architectures of the brain. Empirical evaluation demonstrates significant improvements over linear and spherical baseline models across key metrics—including activation consistency, semantic coherence, and robustness—thereby enhancing interpretability and reproducibility of cross-study findings.

This study investigates the functional role of brain alignment in enhancing the linguistic capabilities of large language models, moving beyond its conventional use as a cognitive modeling tool. To this end, we introduce the first “brain-mismatched” models—architectures deliberately designed to degrade their ability to predict neural activity while preserving core language modeling performance—and systematically compare them against brain-aligned counterparts across more than 200 tasks spanning semantics, syntax, discourse, reasoning, and morphology. Our results demonstrate that impairing brain alignment significantly degrades downstream language task performance, revealing that alignment with human neural representations provides an independent and critical contribution to robust and efficient language understanding. This finding underscores a deep functional link between neurocognitive mechanisms and computational language processing.