NeuroAnalyzer Roadmap

This page describes planned and considered future developments of NeuroAnalyzer. Items are grouped by theme and labeled by status. The roadmap is neither exhaustive nor in a fixed order - priorities shift as the project evolves and as the community provides feedback.

In progress: actively being worked on; Planned: decided and queued; Under consideration: not yet decided.

Have a feature request? Open an issue on Codeberg or GitHub.

Data import & export

  • Planned Full Neurodata Without Borders (NWB) support - TSV/JSON naming, mixed EEG/iEEG/SEEG recordings, STIMULI vs EVENTS handling
  • Planned BIDS-compatible data organization
  • Planned Content-based format detection in import_recording() (magic bytes or header strings) for extensions shared by several formats (.cnt, .eeg, .dat), before adding the formats below
  • Planned Additional import formats, in order of priority:
    • EGI / Magstim MFF (.mff): high-density EEG (128/256 channels), with sensor positions and events
    • Micromed (.trc): clinical EEG and SEEG
    • ANT Neuro eego (.cnt, RIFF container; distinguished from Neuroscan CNT by its content)
    • Compumedics Curry 7/8 (.cdt, .dap, .rs3, .cdt.dpa)
    • BCI2000 (.dat; distinguished from Neuroscan DAT by its content)
    • Artinis fNIRS (.oxy3, .oxy4, .oxy5)
    • Persyst (.lay + .dat), Nihon Kohden (.eeg + .pnt/.21e/.log), Natus/Nicolet (.e)
    • OpenBCI (.txt/.csv), Muse and Emotiv exports
    • MEF3 (iEEG archives, e.g. Mayo, iEEG.org)
    • microelectrode recordings: Blackrock (.ns1–.ns6, .nev), Intan (.rhd, .rhs), Open Ephys (binary), SpikeGLX
    • fNIRS: Hitachi ETG (.csv), Shimadzu (.txt), Cortivision, BOXY (ISS); BIDS-NIRS (_nirs.snirf with channels.tsv and events.tsv)
  • Under consideration MEG formats other than FIFF: CTF (.ds), KIT/Yokogawa (.con, .sqd), BTi/4D

Editing

  • Planned Overlapping fixed-length epochs (epoch(obj; ep_len) with an overlap
  • Planned Insert channel at a given position (add_channel() appends channels at the end)

Processing

  • Planned Automated artifact cleaning: automatic classification of ICA components (correlation with EOG/ECG channels, ICLabel-like classifier), artifact subspace reconstruction (ASR), automatic rejection or interpolation of bad channels and epochs (autoreject-like); available: channel_reject(), epoch_reject(), ica_decompose()/ica_remove(), SSP, artrem_cwd(), denoise_*()
  • Planned Frequency-dependent multitaper smoothing in time-frequency analysis (taper bandwidth per frequency, e.g. wider at higher frequencies; currently one nw/nt for all frequencies)
  • Planned NeuroStim: realistic TES (transcranial electrical stimulation) modeling: electric field in a head model, e.g. import of SimNIBS simulations (see realistic head models in Source imaging); currently a planar point-source model (tes_model())
  • Planned NeuroStim: removal of TES artifacts from EEG (e.g. tACS artifact template subtraction)
  • Planned NeuroStim: TES/TMS protocol plots (protocols: tes_protocol())
  • Under consideration NeuroStim: transcranial focused ultrasound (tFUS)
  • Under consideration NeuroStim: transcranial infrared laser stimulation (TILS)

Analysis

  • In progress Functional connectivity: neural Granger causality (NGC, neural-network-based; classical time-domain and spectral Granger causality: granger(), granger_spectral())
  • Planned Functional connectivity: cross-frequency coupling - phase-amplitude coupling (modulation index, Tort et al.; mean vector length, Canolty et al.), amplitude-amplitude and phase-phase coupling
  • Planned Functional connectivity: orthogonalized amplitude envelope correlation (correction for volume conduction and source leakage, Hipp et al. 2012; amplitude envelope correlation: aecor())
  • Planned Wavelet-based coherence
  • Planned Additional entropy estimators: permutation, approximate, multiscale and spectral entropy (available: histogram, Shannon, log energy, sample and differential entropy, entropy())
  • Planned Additional mutual information estimators: Kraskov-Stögbauer-Grassberger (KSG) estimator, transfer entropy (available: maximum likelihood estimator, mutual_information())
  • Source imaging, in the order of dependence:
    • Planned Realistic head models (BEM): import of head surfaces and head models from FreeSurfer, MNE-Python and SimNIBS, template head (fsaverage); currently spherical head models only
    • In progress Dipole fitting (dipole_fit(); spherical head model) and dipole plots
    • Planned Current density reconstruction (CDR): minimum norm estimate (MNE), sLORETA, eLORETA
    • Planned Beamforming: LCMV, DICS
    • Under consideration Import and segmentation of individual MRI data
  • Planned Interactive ERP tool (ierp(), modeled on ERPLAB)
  • Planned Cluster-based permutation tests (Maris & Oostenveld 2007), e.g. for ERP statistics
  • Planned Integration with the Unfold ecosystem: Unfold.jl (regression ERPs), UnfoldMakie.jl, clusterDepth.jl
  • Under consideration Automated reports (Markdown / HTML / PDF via Pandoc)
  • Under consideration ANOVA test for signal segments

Visualization

  • Planned Egg-shaped (non-circular) head outline (“egghead”) for plot_topo() and plot_locs(), as an alternative to the current circular outline with nose and ears
  • Planned Log frequency scale in plot_psd_3d() (Makie 3D axes have no log scale: log-transformed frequencies with relabeled ticks)
  • Planned ITPC topographical map (plot_itpc_topo(), as plot_psd_topo() and plot_phs_topo())
  • Planned 3D connectome: connections between channels or sources drawn on a 3D head or brain (2D: plot_connectivity_circle(), plot_locs(; connections))
  • Planned Values mapped onto 3D surfaces: scalp topography on a 3D head, source estimates on the cortex (requires the realistic head models, see Source imaging); head and brain meshes: plot_locs3d()

Modality support

  • Planned Improved ECoG (iEEG and SEEG) support, in order of priority:
    • electrode models (grids, strips, depth shafts with contact spacing; shafts parsed from labels, e.g. LA1…LA12); import of contact coordinates (BIDS-iEEG electrodes.tsv, iELVis, LeGUI, 3D Slicer; native or MNI space)
    • montages: automatic bipolar montages along each shaft or strip, common average per shaft or grid, grid Laplacian (neighbors in the grid), white-matter contact reference
    • contact quality: automatic detection of bad contacts (flat signal, line noise, outlying variance or kurtosis, low correlation with neighbors) and of contacts outside the brain
    • epilepsy: interictal epileptiform discharge (spike) detection (e.g. Janča et al. 2015); high-frequency oscillation (HFO) detection, ripples (80–250 Hz) and fast ripples (250–500 Hz) (RMS, Hilbert and MNI detectors); seizure onset zone measures (epileptogenicity index, Bartolomei et al. 2008; line length); cortico-cortical evoked potentials (CCEP): single-pulse stimulation artifact removal, N1/N2 detection, effective connectivity graphs
    • cognitive iEEG: high-gamma (broadband) activity with per-trial baseline normalization (z-score, dB), single-trial high-frequency activity, event-related time-frequency analysis of each contact
    • anatomy and visualization: atlas labels of contacts (DKT, Destrieux, AAL via MNI coordinates); brain-shift correction of grids (projection to the cortical surface, Hermes et al. 2010); values of contacts on a 3D brain surface; contacts on MRI slices
    • group level: MNI normalization, region-of-interest (ROI) aggregation across patients, coverage maps
    • deep brain stimulation (DBS): local field potentials of directional leads, beta burst detection, stimulation artifact removal
    • import formats: Micromed, Nihon Kohden, Natus, MEF3 (see Data import & export), BIDS-iEEG
  • Planned Improved fNIRS support, in order of priority:
    • motion artifact detection (segments with signal changes above a threshold within a short window, as hmrMotionArtifact in Homer3), for rejecting segments or epochs
    • motion correction: CBSI (Cui et al. 2010), spline interpolation (Scholkmann et al. 2010), wavelet filtering (Molavi & Dumont 2012)
    • global signal regression (mean or first principal component of the long channels) for montages without short channels
    • signal quality over time: windowed SCI and peak power (QT-NIRS, Hernandez & Pollonini 2020), marking bad segments; channel pruning by intensity range or SNR (as hmrPruneChannels)
    • GLM: contrasts (e.g. task A − task B); short channels and auxiliary signals (accelerometer, respiration, AUX channels) as regressors (Santosa et al. 2020); HRF temporal and dispersion derivatives and deconvolution bases (FIR, Gaussian); robust weighting (full AR-IRLS); group level (mixed effects or t-test of β values) with FDR correction across channels
    • response measures from block averages (peak amplitude, time to peak, area, mean in a window, SE across epochs) and region-of-interest (ROI) averaging
    • registration to 10-20 or MNI space: labelling channels by the nearest 10-20 position or brain region (fOLD, AAL)
    • tissue oxygenation (StO₂/TOI) from spatially resolved spectroscopy; frequency- and time-domain data
  • Under consideration fNIRS image reconstruction (diffuse optical tomography, DOT): sensitivity matrices and head models
  • Under consideration Eyetracker integration for blink and eye-movement artifact detection
  • Under consideration 3D head scanner support for head mesh (e.g. Polhemus FastSCAN)
  • Under consideration 3D tracker support for electrode localization (e.g. Polhemus G4)

NeuroTester

  • In progress Visual and auditory stimuli presentation module

Performance & infrastructure

  • In progress Further performance optimizations
  • Planned Distributed computing support for analysis pipelines