
Background Mantra recitation is a phonetic-cognitive practice that is thought to engage auditory, phonological, attentional, and motor networks in a graded way across three traditional modes of engagement — passive listening (Shravana), vocal chanting (Vachika Japa), and silent mental recitation (Manasa Japa). Because no synchronized human EEG–fNIRS corpus for the Vishnu Sahasranama (a thousand-name Sanskrit hymn) was available to us, we test the feature-level statistical and classification stages of a proposed multimodal acquisition-and-analysis pipeline on a physiologically parameterized synthetic benchmark dataset rather than on real human recordings; every quantitative result reported below therefore describes the behaviour of the computational pipeline on simulated data, not an empirical finding about human neurocognition. Methods A 90-subject, within-subject benchmark was generated for a three-condition protocol (Listening, Vocal Chanting, Silent Recitation), with EEG (19-channel, 5-band Welch power) and fNIRS (204-channel ΔHbO/ΔHbR) features simulated jointly per synthetic subject from a documented generative model (subject random-effect plus condition effect plus measurement noise; fixed random seed; full parameters and code in the Supplementary Material), yielding a 503-dimensional fused feature vector per condition epoch. Condition effects were compared with repeated-measures ANOVA and Bonferroni-corrected paired-sample post-hoc tests, and three-class discrimination (Listening vs. Vocal Chanting vs. Silent Recitation) was tested with Random Forest and SVM classifiers under subject-grouped, stratified 5-fold cross-validation, separately for EEG-only, fNIRS-only, and fused features. Results In the dataset, repeated-measures ANOVA showed significant condition effects for frontal theta (F(2,178) = 26.06, p < 0.001, partial η2 = 0.226), alpha (F = 46.06, p < 0.001, η2 = 0.341), beta (F = 216.84, p < 0.001, η2 = 0.709), and gamma (F = 19.62, p < 0.001, η2 = 0.181), but not delta (F = 1.45, p = 0.237, ns). Beta power was selectively elevated during Vocal Chanting only (Listening vs. Silent Recitation was non-significant, p = 0.655). Simulated prefrontal HbO differed strongly by condition (F = 335.29, p < 0.001, η2 = 0.790), with a non-monotonic hierarchy of Vocal Chanting (0.0548 ± 0.0128 mM) > Silent Recitation (0.0361 ± 0.0077 mM) > Listening (0.0251 ± 0.0056 mM). Three-class classification on the fused 503-dimensional feature set reached 97.0% (Random Forest) and 95.9% (SVM) mean cross-validated accuracy, a 3.0–2.2 percentage-point gain over the better single-modality model, with prefrontal HbO/HbR features contributing the largest share of Random-Forest discriminative importance (83%). Conclusion Within this the proposed multimodal pipeline recovers internally consistent, condition-specific electrophysiological and hemodynamic patterns and supports above-chance, fusion-benefiting classification of the three engagement modes. These results validate only the feature-level analysis stages (statistics and classification), not the raw-signal acquisition or preprocessing stages, and motivate — but do not yet constitute — human neurocognitive evidence; a human EEG–fNIRS cohort study is required before any claim about actual mantra-chanting neurophysiology can be made.
Background Intracranial hemorrhage (ICH) requires rapid and accurate diagnosis. Artificial intelligence (AI) may support ICH detection on brain CT, particularly where radiological expertise is limited. Objective To assess the diagnostic performance of AI for ICH detection on brain CT and identify factors associated with performance variability. Methods Following PRISMA-DTA recommendations, diagnostic accuracy studies providing sufficient data to construct 2×2 tables were systematically identified. Independent validation cohorts were the primary analytical units, with overlapping datasets excluded from primary pooling (Forty-six studies). Sensitivity and specificity were jointly estimated using a Bayesian bivariate random-effects model. Bayesian subgroup and meta-regression analyses examined validation setting, reference standard, model purpose, and target condition. Results The Bayesian bivariate random-effects model yielded a pooled sensitivity of 91.3% (95% CrI, 89.0–93.3%) and specificity of 95.4% (95% CrI, 93.7–96.6%). The positive and negative likelihood ratios were 19.69 (95% CrI, 14.41–26.99) and 0.091 (95% CrI, 0.070–0.116), respectively, with a diagnostic odds ratio of 216.7 (95% CrI, 131.8–354.3). The sensitivity analysis restricted to eight low-risk-of-bias studies yielded sensitivity of 90.3% (95% CrI, 81.2–95.2%) and specificity of 93.4% (95% CrI, 86.7–96.4%). In separate meta-regression models, expert-based reference standards were associated with higher sensitivity (β = 0.713; 95% CrI, 0.173–1.249; posterior probability, 99.5%), and clinical/real-world validation with higher specificity (β = 0.602; 95% CrI, 0.001–1.175; posterior probability, 97.5%). Conclusion AI showed high diagnostic accuracy for intracranial hemorrhage detection on brain CT, but low-certainty evidence and heterogeneity limit generalizability. Prospective, external, multicenter validation should assess clinical outcomes and workflow impact, including in resource-limited settings.
Microglial activation-induced neuroinflammation is a key process in Parkinson's disease (PD), but the relevant cellular state and the regulatory capacity of gut-derived short-chain fatty acids (SCFAs) remain incompletely resolved. In this study, bulk transcriptomic analysis of the substantia nigra was combined with single-nucleus transcriptomic profiling of the midbrain to examine PD-associated microglial activation and to evaluate the proposed SCFA-HDAC pathway. Pathway-level analysis showed downregulation of dopaminergic and synaptic pathways with concomitant upregulation of neuroinflammatory pathways. Single-nucleus analysis localized this signal to a glycoprotein non-metastatic melanoma protein B (GPNMB)-positive disease-associated microglia (DAM)-like activated subcluster that was consistently enriched in PD donors within this cohort. Using perturbation connectivity analysis and molecular docking, the present computational proxy assessment did not support a uniform reversal of this activation program by HDAC inhibition, and regulation of the core target GPNMB was drug-specific rather than uniform. These findings identify a candidate GPNMB-positive activated microglial phenotype associated with PD neuroinflammation and indicate that support for the proposed SCFA-HDAC axis is limited under the present evidence.
Automated detection of epileptic seizures from scalp electroencephalography (EEG) remains challenging because of strong inter-patient variability. This paper proposes an interpretable, multi-domain feature engineering framework for seizure detection on the CHB-MIT Scalp EEG Database and evaluates it under strictly patient-level protocols. In addition to conventional time-domain, frequency-domain, wavelet and nonlinear descriptors, each EEG segment is passed through a Leaky Integrate-and-Fire neuron, a biologically inspired spike-domain transformation from which eight spike-train statistics are extracted. Gradient-boosted classifiers are evaluated on six entirely unseen test patients, in a 20-patient Leave-One-Subject-Out (LOSO) analysis, and under both a class-balanced and the natural, unbalanced class distribution. XGBoost reaches a ROC-AUC of 0.796 on unseen test patients (0.803 after Optuna tuning) and a mean LOSO ROC-AUC of 0.781, which remains same under the natural class distribution. Paired ablation studies show that every feature domain carries seizure-related information on its own, but that the time-domain family, dominated by peak-to-peak amplitude, is the main source of transferable information: its removal reduces LOSO ROC-AUC, whereas the SNN features, despite a standalone ROC-AUC, add no measurable contribution once conventional features are present. Restricting the positive class to the original seizure annotations increases LOSO ROC-AUC by 0.055, indicating that the extended seizure-associated labelling yields a conservative estimate. Feature-level SHAP analysis identifies amplitude and alpha/beta power markers as the most stable predictors across patients. Together, these results provide a leakage-controlled and interpretable account of what multi-domain EEG features can and cannot contribute to cross-patient seizure detection.