Frontiers in Neuroinformatics最新文献

This study presents a novel fractional order model of Alzheimer's disease (mental disorder) using the Caputo derivative to accurately capture long term memory and hereditary effects in neurodegeneration. The mathematical model incorporates key pathological constituents including neurons, amyloid beta (A _β), tau proteins and microglial responses, allowing detailed simulation of their dynamic interactions. Fundamental properties of the model, including positivity, boundedness, invariant regions and equilibrium points, are rigorously analyzed to ensure biological feasibility. Sensitivity analysis identifies amyloid toxicity as the most influential driver of neuronal loss underscoring its central role in AD progression. Furthermore, a Physics Informed Neural Network (PINN) is developed to approximate system dynamics from noisy observations while ensuring compliance with biological and physical constraints. Compared to standard neural networks the PINN exhibits superior accuracy and robustness especially under data scarcity. By integrating fractional calculus, optimal control and machine learning, this work advances computational modeling of Alzheimer's disease and offers insights into therapeutic optimization.

We developed Macular, a simulation platform with a graphical interface, designed to produce in silico experiment scenarios for the retina and the primary visual system. A scenario involves generating a three-dimensional structure with interconnected layers, each layer corresponding to a type of "cell" in the retina or visual cortex. The cells can correspond to neurons or more complex structures (such as cortical columns). Inputs are arbitrary videos. The user can use the cells and synapses provided with the software or create their own using a graphical interface where they enter the constituent equations in text format (e.g., LaTeX). They also create the three-dimensional structure via the graphical interface. Macular then automatically generates and compiles the C++ code and generates the simulation interface. This allows the user to view the input video and the three-dimensional structure in layers. It also allows the user to select cells and synapses in each layer and view the activity of their state variables. Finally, the user can adjust the phenomenological parameters of the cells or synapses via the interface. We provide several example scenarios, corresponding to published articles, including an example of a retino-cortical model. Macular was designed for neurobiologists and modelers, specialists in the primary visual system, who want to test hypotheses in silico without the need for programming. By design, this tool allows simulation of natural or altered conditions (e.g., pharmacology, pathology, and development).

Introduction: The accumulation of genomic and brain data opens new opportunities for resource friendly, data driven brain exploration. A key challenge is to develop versatile and accessible strategies that integrate and mine multimodal datasets for novel neuroscientific insights. Here, we optimized an integrated workflow for mapping multigenic evolutionary traits in the human brain across cognitive, cellular, and molecular levels.

Methods: At the input stage, the workflow fuses an evolutionary genetic dataset with searchable synthetic functional magnetic resonance imaging (fMRI) databases that are pre clustered into concise psychological domains for improved interpretability. At its core, a Genetic Algorithm for Generalized Biclustering (GABi) mines gene sets under evolutionary selection that also show high expression correlation with fMRI networks.

Results: Applying this workflow, we identified evolutionary patterns spanning cognitive traits, brain cell types, and molecular mechanisms. Focusing on socio affective traits, the algorithm highlighted peaks in adaptive selection in networks for social interaction (language) and social concepts (theory of mind) across hominid, early hominin, and anatomically modern human (AMH) ancestry. These traits emerge from a broad spectrum of excitatory (glutamatergic) and inhibitory (GABAergic) neuronal, as well as non neuronal, cell types. The associated Gene Ontology (GO) terms were enriched for cell signaling, synaptic organization, and neuronal morphology.

Discussion: Together, these findings demonstrate an integrated workflow for molecular to systems level exploration of the brain and provide new perspectives on the evolutionary history of human socio affective functions. This approach can be adapted to screen for functional traits in the context of mental disorders or applied to the brains of other phylogenies in a similar manner.

Background: Spinal cord atrophy is a key biomarker for tracking disease progression in neurological disorders, including multiple sclerosis, amyotrophic lateral sclerosis, and spinal cord injury. Recent MRI advancements have improved atrophy detection, particularly in the cervical region, facilitating longitudinal studies. However, validating atrophy quantification algorithms remains challenging due to limited ground truth data.

Objective: This study introduces SynSpine, a workflow for generating synthetic spinal cord MRI data (i.e., digital phantoms) with controlled levels of artificial atrophy. These phantoms support the development and preliminary validation of spinal cord imaging pipelines designed to measure degeneration over time.

Methods: The workflow consists of two phases: (1) generating synthetic MR images by isolating, extracting and scaling the spinal cord, simulating atrophy on the PAM50 template; (2) performing non-rigid registration to align the synthetic images with the subject's native space, ensuring accurate anatomical correspondence. A proof-of-concept application utilizing the Active Surface and Reg methods implemented in Jim demonstrated its effectiveness in detecting atrophy across various levels of simulated atrophy and noise.

Results: SynSpine successfully generates synthetic spinal cord images with varying atrophy levels. Non-rigid registration did not significantly affect atrophy measurements. Atrophy estimation errors, estimated using Active Surface and Reg methods, varied with both simulated atrophy magnitude and noise level, exhibiting region-dependent differences. Increased noise led to higher measurement errors.

Conclusion: This work presents a novel and modular framework for simulating spinal cord atrophy data using digital phantoms, offering a controlled setting for testing spinal cord analysis pipelines. As the simulated atrophy may over-simplify in vivo conditions, future research will focus on enhancing the realism of the synthetic dataset by simulating additional pathologies, thus improving its application for evaluating spinal cord atrophy in clinical and research contexts.

Background: Timely and accurate assessment of acute ischemic stroke is crucial for determining eligibility for mechanical thrombectomy. The Alberta Stroke Program Early CT Score (ASPECTS) is a widely used tool for evaluating early ischemic changes on non-contrast CT (NCCT), but its interpretation is subject to interobserver variability. Brainomix e-ASPECTS is an automated software designed to standardize and expedite this assessment. We aimed to evaluate the clinical utility and diagnostic performance of the Brainomix e-ASPECTS software in an unselected, real-world cohort of patients undergoing NCCT for suspected acute ischemic stroke.

Methods: We retrospectively analyzed 1,029 NCCT studies from 954 patients between March 2020 and December 2024. e-ASPECTS scores were compared to radiologist-assigned ASPECTS, which served as the reference standard. Diagnostic accuracy, sensitivity, specificity, and correlation between scoring methods were assessed.

Results: There was a strong correlation between e-ASPECTS and radiologist ASPECTS (ρ = 0.953, p < 0.001). For detecting acute ischemia, sensitivity was 95.8% (95% CI, 93.6-97.3%), specificity 96.9% (95% CI, 94.7-98.2%), and overall accuracy 96.3% (95% CI, 95.1-97.5%). The positive predictive value was 97.2% (95% CI, 95.3-98.4%), and the negative predictive value was 95.3% (95% CI, 92.8-96.9%). Score concordance was high, with exact matches in 92.3% of cases and a ≤ 1-point difference in 97.7%. Misclassification for thrombectomy eligibility (ASPECTS < 6) occurred in four cases (0.4%). The software achieved a processing success rate of 91.9%.

Conclusion: E-ASPECTS demonstrates high diagnostic accuracy and strong agreement with expert radiological assessment, supporting its role as a valuable decision support tool in acute stroke imaging. However, its use should complement, not replace, expert interpretation, particularly in patients with low ASPECTS scores, where treatment decisions are most sensitive.

Dementia poses a major challenge to individuals and public health systems. Detecting cognitive decline through spontaneous speech offers a promising, non-invasive avenue for diagnosis of mild cognitive impairment (MCI) and dementia, enabling timely intervention and improved outcomes. This study describes our submission to the PROCESS Signal Processing Grand Challenge (ICASSP 2025), which tasked participants with predicting cognitive decline from speech samples. Our method combines eGeMAPS features from openSMILE, HuBERT (a self-supervised speech representation model), and GPT-4o, OpenAI's state-of-the-art large language model. These are integrated with the custom LSTM and ResMLP neural networks, and supported by Scikit-learn regressors/classifiers for both cognitive score regression and dementia classification. Our regression model based on LightGBM achieved an RMSE of 2.7775, placing us 10th out of 80 teams globally and surpassing the RoBERTa baseline by 7.5%. For the three-class classification task (Dementia/MCI/Control), our LSTM model obtained an F1-score of 0.5521, ranking 20th of 106 and marginally outperforming the best baseline. We trained models on speech data from 157 study participants, with independent evaluation performed on a separate test set of 40 individuals. We discoved that integrating large language models with self-supervised speech representations enhances the detection of cognitive decline. The proposed approach offers a scalable, data-driven method for early cognitive screening and may support emerging applications in neuropsychological informatics.

Alzheimer's disease (AD) and frontotemporal dementia (FTD) are major neurodegenerative disorders with characteristic EEG alterations. While most prior studies have focused on eyes-closed (EC) EEG, where stable alpha rhythms support relatively high classification performance, eyes-open (EO) EEG has proven particularly challenging for AD, as low-frequency instability obscures the typical spectral alterations. In contrast, FTD often remains more discriminable under EO conditions, reflecting distinct neurophysiological dynamics between the two disorders. To address this challenge, we propose a CNN-based framework that applies Dynamic Mode Decomposition (DMD) to segment EO EEG into shorter temporal windows and employs a 3D CNN to capture spatio-temporal-spectral representations. This approach outperformed not only the conventional short-epoch spectral ML pipeline but also the same CNN architecture trained on FFT-based features, with particularly pronounced improvements observed in AD classification. Excluding delta yielded small gains in AD-involving contrasts, whereas FTD/CN was unchanged or slightly better with delta retained-suggesting delta is more perturbative in AD under EO conditions.

Introduction: Autism Spectrum Disorder (ASD) diagnosis remains complex due to limited access to large-scale multimodal datasets and privacy concerns surrounding clinical data. Traditional methods rely heavily on resource-intensive clinical assessments and are constrained by unimodal or non-adaptive learning models. To address these limitations, this study introduces AutismSynthGen, a privacy-preserving framework for synthesizing multimodal ASD data and enhancing prediction accuracy.

Materials and methods: The proposed system integrates a Multimodal Autism Data Synthesis Network (MADSN), which employs transformer-based encoders and cross-modal attention within a conditional GAN to generate synthetic data across structural MRI, EEG, behavioral vectors, and severity scores. Differential privacy is enforced via DP-SGD (ε ≤ 1.0). A complementary Adaptive Multimodal Ensemble Learning (AMEL) module, consisting of five heterogeneous experts and a gating network, is trained on both real and synthetic data. Evaluation is conducted on the ABIDE, NDAR, and SSC datasets using metrics such as AUC, F1 score, MMD, KS statistic, and BLEU.

Results: Synthetic augmentation improved model performance, yielding validation AUC gains of ≥ 0.04. AMEL achieved an AUC of 0.98 and an F1 score of 0.99 on real data and approached near-perfect internal performance (AUC ≈ 1.00, F1 ≈ 1.00) when synthetic data were included. Distributional metrics (MMD = 0.04; KS = 0.03) and text similarity (BLEU = 0.70) demonstrated high fidelity between the real and synthetic samples. Ablation studies confirmed the importance of cross-modal attention and entropy-regularized expert gating.

Discussion: AutismSynthGen offers a scalable, privacy-compliant solution for augmenting limited multimodal datasets and enhancing ASD prediction. Future directions include semi-supervised learning, explainable AI for clinical trust, and deployment in federated environments to broaden accessibility while maintaining privacy.