Key concepts

Neurology in Gradient Biotech is organized around brain networks, electrophysiology recordings, functional imaging, cohort comparisons, and reproducible analysis runs. This page explains the neuroscience concepts behind the workflows and the app concepts used to manage them.

Brain connectivity data

Many Neurology workflows start from a connectivity matrix: a square table where each row and column represents a brain region, sensor, or node, and each value represents the strength of a connection between two nodes.

Common matrix meanings:

• Correlation matrix: connection values represent statistical association between regional time series.
• Partial correlation matrix: connection values estimate association after accounting for other regions.
• Binary adjacency matrix: values indicate whether an edge is present or absent.
• Weighted adjacency matrix: values represent connection strength.

A connectivity matrix is a model of a brain network, not the brain itself. Interpretation depends on how regions were defined, how signals were preprocessed, how edges were thresholded, and whether the matrix represents structural, functional, or electrophysiology-derived connectivity.

Graph theory and network metrics

Graph theory treats brain regions or sensors as nodes and connections as edges. Network metrics summarize the topology of those edges.

Common graph concepts:

Concept	Meaning
Node	Brain region, electrode, sensor, or network unit.
Edge	Connection between two nodes.
Degree	Number of edges connected to a node.
Strength	Sum of weighted connections for a node.
Clustering	Tendency for a node's neighbors to connect with each other.
Path length	Number or cost of steps needed to travel through the network.
Global efficiency	How efficiently information can move across the network.
Modularity	Degree to which the network separates into communities or modules.
Small-worldness	Balance of local clustering and short global paths compared with random networks.
Betweenness centrality	How often a node lies on shortest paths between other nodes.

The graph metrics pipeline uses Brain Connectivity Toolbox (BCT)-equivalent measures. These outputs are useful for research comparison, but they are sensitive to thresholding, density, parcellation, and preprocessing choices.

Network visualizations

Connectivity results can be inspected through multiple views:

• Heatmap: matrix view where color represents connection strength.
• Network graph: node-edge layout, often colored by module assignment.
• Three-dimensional brain view: region positions shown in brain space when atlas coordinates are available.

Visualization helps with quality control and hypothesis generation. It should be paired with numeric outputs and methods details before making biological claims.

Electroencephalography

Electroencephalography (EEG) records electrical activity from electrodes on the scalp. It is used to study oscillations, event-related activity, connectivity, sleep, attention, seizure-related patterns, and brain-state dynamics.

Key electroencephalography concepts:

• Channel: one electrode or signal trace.
• Sampling rate: number of signal samples recorded per second.
• Bandpass filter: keeps frequencies inside a target range.
• Notch filter: suppresses narrow noise, often line noise.
• Re-reference: changes the voltage reference used for channels.
• Bad channel: noisy or unusable channel flagged for repair or exclusion.
• Independent component analysis (ICA): decomposition method often used to identify artifact components.

Electroencephalography preprocessing prepares the recording for feature extraction by filtering, handling bad channels, re-referencing, and optionally applying artifact rejection.

Electroencephalography features

After preprocessing, the app can compute spectral and connectivity features from electroencephalography recordings.

Common feature concepts:

• Band power: signal power in frequency bands such as delta, theta, alpha, beta, and gamma.
• Spectral entropy: summary of how distributed or concentrated the frequency content is.
• Phase-locking value (PLV): phase consistency between two signals.
• Coherence: frequency-specific coupling between two signals.
• Weighted phase lag index (wPLI): phase-lag connectivity metric designed to reduce zero-lag volume-conduction effects.

Electroencephalography-derived connectivity matrices can feed graph metrics, allowing sensor or source-level networks to be summarized with the same graph concepts used for other connectomes.

Microstates and time-frequency analysis

Microstates are short, recurring scalp topography patterns in electroencephalography data. They are often interpreted as brief brain-state configurations.

Microstate outputs can include:

• State map: scalp pattern for one microstate.
• Coverage: fraction of time assigned to a state.
• Duration: average length of state visits.
• Occurrence: how often a state appears.
• Transitions: movement between states over time.
• Global explained variance (GEV): how much variance the microstate model explains.

Time-frequency analysis shows how spectral power changes over time. It is useful for studying oscillatory dynamics, task responses, and changing brain states.

Functional magnetic resonance imaging

Functional magnetic resonance imaging (functional MRI or fMRI) measures blood oxygen level-dependent signal changes that are used as an indirect marker of neural activity.

Common functional imaging concepts:

• Blood oxygen level-dependent (BOLD) signal: imaging signal related to blood oxygenation changes.
• NIfTI: Neuroimaging Informatics Technology Initiative file format for imaging data.
• Atlas parcellation: assignment of brain voxels to named regions of interest.
• Region of interest (ROI): brain region used to extract a time series.
• Confound regression: removal of nuisance signals such as motion or physiological artifacts.
• Pearson correlation: correlation between region time series.
• Fisher z transformation: transformation often applied to correlations before statistical comparison.

The functional connectivity pipeline extracts regional time series, applies preprocessing choices such as confound regression and bandpass filtering, and produces a connectivity matrix for graph analysis.

Brain Imaging Data Structure

Brain Imaging Data Structure (BIDS) is a standardized folder and metadata convention for neuroimaging datasets. BIDS validation checks whether uploaded archives include expected files, subject/session structure, and metadata.

Validation does not prove the science is correct. It checks structure so downstream functional connectivity workflows have the metadata they need.

Dynamic functional connectivity

Dynamic functional connectivity estimates how connectivity patterns change over time. The app supports sliding-window correlation and state identification.

Important concepts:

• Sliding window: time window moved across a recording to estimate changing connectivity.
• Connectivity state: recurring connectivity pattern identified from windows.
• State timecourse: sequence of state assignments over time.
• Transition statistic: summary of movement between states.

Dynamic connectivity results are sensitive to window length, recording duration, noise, and the number of states chosen.

Cohorts and Network-Based Statistics

A cohort is a set of subjects or datasets compared across groups. In Neurology, cohorts often compare patient versus control groups, treatment arms, symptom strata, or task conditions.

Network-Based Statistics (NBS) is a group-comparison method for connectivity matrices. It tests whether connected components of edges differ between groups while controlling for multiple comparisons at the network-component level.

Key cohort concepts:

• Group label: category used for comparison.
• Batch graph metrics: running graph metrics across many subject matrices.
• Effect size: magnitude of a group difference.
• Permutation testing: repeated label shuffling used to estimate statistical significance.
• Significant component: connected set of edges with group differences passing the chosen threshold.

Network-Based Statistics results should be interpreted with group size, preprocessing consistency, matrix density, threshold choices, and cohort metadata.

Artificial intelligence interpretation

Artificial intelligence interpretation summarizes completed metrics with explicit citations to run artifacts.

Interpretation contexts include:

Context	When	Output
Graph metrics	Single matrix analysis complete	Network topology summary with cited Brain Connectivity Toolbox-equivalent values
Network-Based Statistics	Cohort comparison with significant components	Group-difference narrative with edge statistics
Cohort metrics	Batch graph metrics across groups	Group-level metric comparison summary
Microstates	Electroencephalography microstate segmentation complete	State coverage and transition interpretation
Prediction	Cross-validated classification or regression run complete	Model performance and stable-edge narrative

Interpretations do not compute new metrics and do not diagnose neurological or psychiatric conditions. They explain completed outputs and include a research disclaimer.

Group-level comparison of microstate statistics (coverage, duration, occurrence, GEV between cohorts) is not yet available — cohort comparison today covers connectivity edges (NBS) and global graph metrics only.

Study

A study is the top-level container for one neuroscience project. It holds datasets, subject records, pipeline runs, and cohort analyses. Create one study per protocol, imaging cohort, or analysis project.

Each study has a primary modality, such as connectivity matrix, electroencephalography, functional magnetic resonance imaging, or multimodal data. The modality guides default workflows.

Dataset

A dataset is one uploaded file or recording attached to a study. Each dataset records:

• Modality, such as connectivity matrix, electroencephalography, functional magnetic resonance imaging, or multimodal data
• File format, such as comma-separated values (CSV), NumPy archive (NPZ), European data format (EDF), Neuroimaging Informatics Technology Initiative (NIfTI), or Brain Imaging Data Structure archive
• Detected metadata, such as region count, channel count, or sampling rate
• Optional group label for cohort comparison
• Storage path on the local filesystem

Open a dataset to run pipelines, explore visualizations, and view run history.

Subject

A subject represents one participant. Subjects have an external identifier plus optional demographics, diagnosis, group label, cognitive scores, medications, and flexible metadata imported from comma-separated values.

Link datasets to subjects when cohort metadata should align across modalities.

Pipeline run

A run is one execution of an analysis pipeline.

Each run stores:

• Unique run identifier
• Pipeline type and version
• JavaScript Object Notation (JSON) parameter record
• Status: queued -> running -> complete or failed
• Output artifacts, such as JSON metrics, matrices, overlays, and interpretations
• Timestamps and duration

Runs chain through provenance. Graph metrics reference the source matrix, electroencephalography features reference their preprocess run, functional connectivity references the source imaging dataset, and Network-Based Statistics runs reference batch graph-metric outputs.

Job polling

Every analysis runs in the background. Status badges update automatically on dataset and cohort pages as jobs progress, so you don't need to wait on a loading screen.

Artifacts

Pipeline outputs are stored as structured data tied to the run.

Examples:

• metrics.json: global and per-node Brain Connectivity Toolbox-equivalent graph metrics
• connectivity_matrix.json: processed matrix for visualization
• eeg_features.json: band power and connectivity summaries
• nbs_report.json: significant edge components from cohort Network-Based Statistics
• interpretation.json: artificial intelligence interpretation output

Download artifacts through the metadata API, or access them directly on the filesystem in local development.

Provenance and reproducibility

Every analysis should be reproducible. Open the Runs page for a study to review the full run history — pipeline type, status, and timestamps — for every completed, running, and failed job. Cite the run_id and pipeline_version from a run record in your own methods text; there is no dedicated methods-text generator, CSV/manifest export, annotation, or one-click re-run UI today.

Interpretation scope

The Neurology area supports exploratory neuroscience research workflows. It is not a regulated clinical diagnostic or neurological assessment tool. Treat outputs as research metrics for quality control, hypothesis testing, and cohort analysis.