Almudena Capilla

Neural oscillations are recognized as a fundamental component of brain electromagnetic activity. They are implicated in a wide range of cognitive processes and proposed as a core mechanism for brain communication. Nonetheless, detecting genuine neural oscillations remains a methodological challenge, particularly due to the difficulty of distinguishing them from aperiodic background activity. To identify episodes of oscillatory activity directly at their sources, we developed sBOSC, which extends the BOSC (Better OSCillation detection) family of algorithms. Consistent with existing approaches, sBOSC detects oscillatory episodes that exceed both a defined power threshold and a minimum duration criterion. In sBOSC, however, the detection of oscillatory episodes also relies on identifying peaks (i.e., local maxima) in the power spectra as well as throughout the brain volume (spatial peaks). Using a series of simulated signals, we tested the ability of sBOSC to detect and localize oscillations across multiple scenarios. Our results show that most oscillatory episodes were accurately detected at their sources, achieving above 95% accuracy under optimal conditions (i.e., high signal‐to‐noise ratio, lower frequencies, and numerous successive cycles). In addition, we validated sBOSC's performance using real magnetoencephalography (MEG) data from both resting‐state and motor task recordings. From the detected oscillatory episodes, we extracted a topographic distribution of natural frequencies that is consistent with previous work, as well as the expected alpha‐ and beta‐band modulations over sensorimotor regions during motor preparation. In conclusion, sBOSC offers a novel approach for identifying oscillatory activity in electrophysiological signals. It extends previous algorithms by operating in source space and verifying the presence of genuine spectral peaks, thereby enabling new possibilities for exploring brain dynamics.

Most of what we currently know about brain oscillations is derived from Fourier-based spectral power methods. While widely adopted, these procedures introduce methodological limitations and inherently overlook informative neurophysiological features that often remain unreported. In this study, we characterized resting-state oscillatory activity from human magnetoencephalography (MEG) recordings (N = 128) by integrating two complementary approaches. First, oscillatory episodes were detected at the source level with sBOSC. Subsequently, the ByCycle algorithm was applied to these episodes to extract individual cycle features. Results revealed that the brain engages in oscillatory activity for only ∼25% of the recording time, with occipital and parietal regions accounting for the highest temporal prevalence across canonical frequency bands. Furthermore, oscillatory episodes lasted an average of 4.6 cycles, reinforcing the view of neural oscillations as transient bursts. Region-specific duration and power measures revealed distinct anatomical organizations offering complementary physiological information. Finally, by extracting the instantaneous amplitude, period, and waveform asymmetry of individual cycles, we successfully dissociated sinusoidal occipital alpha waves from the asymmetric sensorimotor mu rhythm. By moving beyond traditional power-centric analyses, this approach provides a comprehensive characterization of spontaneous oscillatory activity, thereby offering new insights into the spatial, temporal, and spectral structure of human brain oscillations.

Neural oscillations are fundamental to brain function and cognition. Conventional analyses often rely on predefined frequency bands to assess power modulations, which may obscure finer-grained spectral variability. In this study, we focused on frequency rather than power to investigate whether the natural frequency of each brain region, typically observed at rest, represents a stable intrinsic property or dynamically reconfigures during cognitive processing. We analysed magnetoencephalography (MEG) data from the Human Connectome Project (HCP) across motor execution, working memory, and language processing tasks. Using a multivariate, data-driven spectral clustering approach, we mapped natural frequencies on a voxel-by-voxel basis without imposing predefined bands or regional boundaries. Results indicated that, while the global spatial organization of natural frequencies remained largely preserved during task engagement, specific cortical regions exhibited systematic, task-dependent shifts. In the sensorimotor cortices, the typical resting frequency of ∼24 Hz decreased to ∼6 Hz during movement preparation and at movement onset, and shifted to high-beta rhythms (∼30 Hz) following hand movement. Increased working memory demands accelerated parieto-occipital alpha/beta activity (from ∼11/16 Hz to ∼13/20 Hz) and recruited high-gamma oscillations (60 to 80 Hz) in medial temporal regions. Finally, arithmetic processing elicited a ∼5 to 15 Hz increase within the beta/gamma ranges across frontoparietal networks relative to semantic comprehension. Taken together, these findings demonstrate that natural frequencies reflect a hybrid architecture: globally stable, yet locally flexible in response to cognitive demands. Moreover, our results suggest that cognitive engagement tends to accelerate neural rhythms in functionally specialized regions, providing a more nuanced understanding of the spectral architecture of human brain function beyond conventional power- and band-based metrics.

Neural oscillations are central to brain function and communication, yet they are typically characterized in terms of spectral power within predefined frequency bands, potentially obscuring their underlying functional organization. An alternative framework focuses on oscillatory frequency rather than power, revealing that each brain region exhibits a characteristic, or natural, frequency that can be estimated at the voxel level using a data-driven approach. Although this framework has been successfully applied to MEG, its broader use remains limited by cost and availability. Here, we extended this approach to EEG and validated it against MEG-derived maps, assessing its robustness across EEG channel densities (high-density, 64 channels; low-density, 32 channels) and physiological states (eyes open and closed). EEG-derived maps revealed a coherent spatial organization of natural frequencies across the cortex, reproducing the large-scale posterior-to-anterior and medial-to-lateral gradients of increasing frequency previously described with MEG. Differences between MEG and EEG were mainly confined to frontal and temporal regions, likely reflecting the differential sensitivity of the two techniques to neural source configurations, whereas posterior regions showed highly similar patterns. Importantly, this organization remained stable despite reductions in EEG sensor density and was modulated by physiological state, reproducing the well-known posterior alpha dominance during eyes-closed conditions. Together, these findings demonstrate that natural frequency mapping can be extended beyond specialized MEG research environments to low-density EEG settings, offering an accessible and scalable tool for investigating brain oscillations and their alterations in neuropsychiatric conditions.

Selective internal attention allows working memory (WM) contents to be prioritized according to task demands, a process typically investigated with the retro-cue paradigm. In this paradigm, a cue presented during WM maintenance indicates which item will be tested, enabling voluntary allocation of attentional resources and typically enhancing memory performance. Recent findings suggest, however, that saccade preparation toward the location of a WM item can also produce an involuntary prioritization effect. In the present study, we compared behavioral performance and neural oscillatory dynamics associated with involuntary prioritization induced by saccade preparation and voluntary prioritization elicited by spatial and color retro-cues. Behaviorally, saccade-based prioritization produced only a modest benefit relative to incongruent items, whereas both types of retro-cues yielded substantially larger improvements. At the neural level, alpha-band lateralization showed a similar spatial-attention pattern across all prioritization conditions, supporting a shared covert attentional mechanism. In contrast, bilateral posterior alpha power was reduced following retro-cues compared to saccadic preparation, consistent with a release of WM load when non-cued items are deprioritized. Overall, our results indicate that voluntary and involuntary mechanisms of WM prioritization rely on overlapping spatial-attention processes, but differ in strategic control and their consequences for load management. Importantly, top-down prioritization through retro-cues produces a stronger behavioral enhancement than involuntary prioritization driven by saccade preparation.

In this study, we investigated whether individual brain maps of natural frequencies derived from EEG can serve as physiologically meaningful biomarkers for Parkinson’s disease (PD), a disorder characterized by altered oscillatory activity. Data were gathered from three independent, openly available EEG databases. Natural frequency maps were extracted for 57 PD participants and 57 sex- and age-matched healthy controls (HC). The group-level brain maps showed region-specific slowing of oscillatory activity in PD, particularly in frontomedial and left frontolateral cortices, as well as hemispheric asymmetries consistent with the lateralized onset of motor symptoms. A subset of 100 participants (50 PD, 50 matched controls) was then used to train and internally validate three machine learning classifiers—support vector machine (SVM), random forest, and logistic regression—under different PCA-based dimensionality reduction and feature normalization schemes. Among all models, SVM achieved the best performance (AUC ≈ 77%). This classifier was subsequently retrained on the full set of 100 participants and evaluated on an independent hold-out test set of 14 participants, achieving an AUC of 75.5% with 71.4% accuracy. Feature contribution analysis highlighted the regions that were most informative for classification toward each class, complementing the group-level observations. Overall, these results indicate that natural frequency mapping captures disease-related alterations in cortical dynamics and provides interpretable features for EEG-based diagnostic tools, opening new avenues for biomarker development in oscillopathies.

Neural oscillations are recognized as a fundamental component of brain electromagnetic activity. They are implicated in a wide range of cognitive processes and proposed as a core mechanism for brain communication. Nonetheless, detecting genuine neural oscillations remains a methodological challenge, particularly due to the difficulty of distinguishing them from aperiodic background activity. To identify episodes of oscillatory activity directly at their sources, we developed sBOSC, which extends the BOSC (Better OSCillation detection) family of algorithms. Consistent with existing approaches, sBOSC detects oscillatory episodes that exceed both a defined power threshold and a minimum duration criterion. In sBOSC, however, the detection of oscillatory episodes also relies on identifying peaks (i.e., local maxima) in the power spectra as well as throughout the brain volume (spatial peaks). Using a series of simulated signals, we tested the ability of sBOSC to detect and localize oscillations across multiple scenarios. Our results show that most oscillatory episodes were accurately detected at their sources, achieving above 95% accuracy under optimal conditions (i.e., high signal-to-noise ratio, lower frequencies, and numerous successive cycles). In addition, we validated sBOSC’s performance using real magnetoencephalography (MEG) data from both resting-state and motor task recordings. From the detected oscillatory episodes, we extracted a topographic distribution of natural frequencies that is consistent with previous work, as well as the expected alpha- and beta-band modulations over sensorimotor regions during motor preparation. In conclusion, sBOSC offers a novel approach for identifying oscillatory activity in electrophysiological signals. It extends previous algorithms by operating in source space and verifying the presence of genuine spectral peaks, thereby enabling new possibilities for exploring brain dynamics.

Neural oscillations are critical for brain function and cognition. Thus, identifying the typical or natural frequencies of the brain is an important step in understanding its functional architecture. Recently, a data-driven algorithm has been developed for mapping these frequencies throughout the cortex, free of anatomical and frequency-band constraints, but its robustness is limited to group-level analyses. Here, we adapt this algorithm to improve the single-subject maps of natural frequencies derived from magnetoencephalography and validate them using the fingerprinting technique. Modifications to the original method included (1) increasing the number of power spectra assigned to each k-means cluster, and (2) smoothing across neighboring voxels. Our results show high accuracy in individual identification within single sessions and across sessions separated by over four years. This demonstrates the stability and reliability of the single-subject mapping of natural frequencies, enhancing opportunities for identification of pathological variations in the intrinsic oscillatory activity of individuals.