Early differentiation of mild cognitive impairment (MCI) from Alzheimer’s disease (AD) remains difficult because diagnosis integrates cognitive, imaging, laboratory, and neurological evidence. EEG-based models may provide complementary information, but fixed descriptors can miss signal-specific structure. SOFEx was developed as an adaptive and interpretable EEG feature-engineering framework. The analysis used 1,905 EEG segments from 451 recordings of 168 participants (59 AD, seven MCI, and 102 controls). SOFEx combined self-organized rose pattern extraction, fold-wise feature selection, channel-wise classification, and iterative majority voting. Validation used leave-one-subject-out (LOSO), leave-one-record-out (LORO), and participant-grouped 10-fold cross-validation (CV). Five classifiers were compared under LOSO, and a support vector machine (SVM) was retained. LOSO, used as the primary evaluation of generalization to unseen participants, achieved 84.09
High-altitude hypoxia has been examined with conventional electroencephalography (EEG), event-related potentials, and neuroimaging, but its association with millisecond-scale resting-state EEG microstate dynamics and peripheral physiology remains unclear. We studied healthy adults including 124 low-altitude (LA) participants (50 men, 74 women) and 174 high-altitude (HA) participants (42 men, 132 women; age 28.05 ± 8.59 years; age unavailable for LA). Coverage, duration, global explained variance (GEV), occurrence, and transition probabilities were extracted for four canonical microstates. LA–HA differences were tested using unadjusted linear models; within-HA associations among the high-altitude acclimatization index (AAI), serum albumin, and Microstate A-related indicators were examined using covariate-adjusted regressions and exploratory bootstrapped mediation analyses. Compared with LA participants, HA participants showed increased Microstate A duration and GEV, decreased occurrence, and unchanged coverage, together with reduced Microstate B-related indicators, enhanced Microstate C-related indicators, and increased A→B and A→C transitions. Within HA participants, AAI was negatively associated with serum albumin, and exploratory mediation analyses indicated indirect associations between AAI and Coverage A, Occurrence A, GEV A, and Microstate A-originating transitions through serum albumin. These findings suggest that chronic high-altitude exposure is associated with resting-state EEG microstate reorganization in healthy adults and that acclimatization-related individual differences may be linked to Microstate A dynamics through serum albumin-related peripheral status.
Resting-state EEG studies in Parkinson’s disease (PD) have largely emphasized static power or connectivity, whereas fixed-sequence OFF-to-ON session differences may also involve the temporal organization of network states. Paired source-space evidence under such fixed acquisition protocols remains limited. We analyzed 64-channel eyes-closed resting-state EEG in 39 patients with PD contributing paired OFF/ON recordings for the prespecified 90 s primary analysis; 24 also contributed paired 178 s recordings for duration robustness. After individual MRI-based source reconstruction, low-beta (13–20 Hz) dynamic amplitude-envelope connectivity was computed with 2 s windows. Primary endpoints were switching rate (SR) and normalized Shannon entropy (H); fractional occupancy (FO_path) and mean dwell time (MDT_path) of an internally defined OFF-dominant state were prespecified secondary conditional endpoints and are interpreted only as conditional summaries. PD-only linear mixed-effects models quantified the fixed-sequence OFF-to-ON session contrast; medication state and session order were necessarily collinear. In the 90 s analysis, the later ON session showed higher SR (β = 3.3224, Holm p = 0.00406) and H (β = 0.06781, Holm p = 0.000749) than the earlier OFF session. FO_path (β = −0.09167, Holm p = 1.08 × 10⁻⁵) and MDT_path (β = −1.05171, Holm p = 0.00524) were lower in the later ON session. The 178 s reanalysis preserved the same directions, although SR and especially MDT_path were more parameter-sensitive across sensitivity analyses. The fixed OFF-then-ON protocol was associated with reproducible within-subject differences in low-beta state dynamics, with the strongest evidence arising from SR and H. Because medication state was perfectly collinear with session order, these findings cannot distinguish medication-specific effects from order-, time-, vigilance-, fatigue-, or adaptation-related influences. FO_path and MDT_path remain conditional summaries tied to an internally defined OFF-dominant state, and MDT was less stable in the 178 s and OFF-only sensitivity analyses.
Epilepsy is one of the most heterogeneous and unpredictable neurological disorders, characterized by complex brain dynamics, significant interindividual variability, and a therapeutic response that is often difficult to predict. Despite major advances in neuroimaging, electrophysiology, and antiepileptic treatments, clinical decision-making still faces major challenges, including precisely localizing the epileptogenic zone, predicting seizures, identifying drug-resistant patients, and optimizing personalized therapeutic strategies. Over the past two decades, computational models have gradually emerged as promising tools for addressing these challenges. Early approaches relied primarily on mechanistic biophysical models aimed at reproducing the neural dynamics and mechanisms of hyperexcitability involved in epileptogenesis. More recently, the rise of artificial intelligence and deep learning has facilitated the development of predictive models capable of leveraging large-scale multimodal data. Finally, the emergence of digital brain twins and adaptive closed-loop systems now paves the way for personalized computational medicine integrating simulation, prediction, and real-time therapeutic intervention. This review provides a critical analysis of the major international initiatives developed between 2004 and 2026, including the EPICLUSTER/EBRAINS, EPINOSTICS, ENIGMA-Epilepsy, NeuroPace RNS, SeizeIT, EpiBioS4Rx, and NeuroVista projects, as well as several translational platforms based on artificial intelligence and biophysical modeling. We evaluate the actual contribution of these models to clinical decision-making in epilepsy, their translational contributions, their methodological limitations, and the challenges associated with their integration into hospital settings. We argue that the future of computational epileptology does not rest on the opposition between artificial intelligence and mechanistic models, but on their convergence within explainable, personalized, and clinically interpretable hybrid architectures.
The thalamus is critical relay hub of brain connectivity, yet its topographical organization during the early human brain development remains incompletely understood. Our study characterized the critical perinatal developmental period by optimizing and applying a clustering framework to parcellate the neonatal thalamus and delineate its developmental trajectory. Thalamus topography was studied using Gaussian Mixture Model (GMM) based clustering applied to voxel-level fiber orientation distribution functions, obtained from diffusion MRI data. Three perinatal datasets were used: term-born infants from developing human connectome project (dHCP, n = 68, age at MRI: 37.4–44.7 weeks), dHCP preterm infant data (n = 48, 29.3–37.4 weeks) and a neonatal control dataset from Zürich (n = 42, 39.0-48.7 weeks). Clustering was optimized using an adult test-retest diffusion MRI dataset (n = 30). We subdivided the thalamus into seven clusters, where GMM showed moderate reproducibility (test-retest overlap ranged from 0.65 to 0.73), and comparable anatomical alignment with digital Morel atlas. Robust regression analysis revealed significant age-related volume increases in absolute volumes across thalamic clusters (FDR-corrected p < 0.001). Topography by means of relative volumes remained stable, suggesting preservation of thalamic topography during late gestation. Two clusters, closely corresponding to the ventral-anterior nucleus and pulvinar group, showed moderate age-related topographical changes (FDR-corrected p < 0.05, adjusted R²=0.039 and 0.101). Group-wise analysis of harmonized volumes revealed no significant group effects between preterm and term born neonates. No sex differences were observed. Our findings demonstrate that thalamic topography is largely established across the perinatal period; however, the lack of systematic term-equivalent scans in preterm infants limits our ability to draw conclusions about the effects of prematurity. The interpretation might be limited by the indirect nature of diffusion MRI features and the moderate reproducibility of the clustering techniques.
