Electroencephalography (EEG) is a relatively inexpensive and practical way to record brain activity with millisecond resolution. That combination makes it possible to study changes in brain activity during an experiment, at the bedside, or while someone operates a device. The difficulty is that movement, muscle activity, and overlapping responses can obscure the features an analysis depends on. In addition, recovering a feature is only the first step. Its interpretation depends on how the study was designed, and its practical value depends on how well it performs under the conditions in which it will be used. This Special Issue titled “Advancing Signal Processing and Analytics of EEG Signals” brings together eight research articles, four reviews, and one case report that examine these questions across a range of EEG applications.
1. Preserving the EEG Feature of Interest
Reducing artifacts starts with how EEG is recorded. In ten healthy volunteers performing standing tasks, Ivaldi et al. [1] found preliminary evidence that a cap with neoprene-jacketed electrodes, used with a quantitative processing workflow, reduced movement-related contamination. Once the signal is recorded, the challenge is to remove artifacts without losing the activity of interest. Grobbelaar et al. [2] reviewed wavelet methods with this trade-off in mind, including which methods can operate online, automatically, or with a single channel. Versaci and La Foresta [3] examined how fuzzy methods handle cases in which neural activity and artifact are difficult to distinguish.
Processing also shapes how we describe the activity we recover. Arrufat-Pié et al. [4] used multivariate empirical mode decomposition to study EEG from 34 healthy volunteers without imposing fixed frequency bands. The Hilbert-Huang analysis produced less overlap between two adjacent modes than the Fourier-based analysis, although the interpretation of mode mixing in the alpha and theta ranges remains open.
2. From Waveforms to Candidate Markers
Detecting a waveform and explaining its significance require different evidence. Abbasi et al. [5] found sharp waves of similar shape in EEG from seven preterm fetal sheep after hypoxia-ischemia and seven extremely preterm infants monitored within six hours of birth. In a follow-up study, a wavelet-scalogram classifier performed best at detecting these patterns, with other methods close enough for computational cost to influence the choice [6]. Whether the detected waves add prognostic information beyond routine EEG assessment could be tested prospectively against independently assessed outcomes. Tozzi [7] also began with waveform shape, asking whether some pathological transients behave as solitons. That hypothesis could be tested by checking whether the waves retain their shape as they propagate.
A similar question arises when the feature is a relationship between signals. Using Bayesian estimates of sensor-space connectivity, Gangapuram and Manian [8] distinguished rest from mental arithmetic with a mean within-participant accuracy of 91% in the beta band across 36 participants. Graph measures also differed between groups defined by task performance. The study shows that connectivity estimates can distinguish task states under the study protocol. Attributing a change in connectivity to an intervention requires a different design. Muroni et al. [9] reported a change in estimated frontotemporal interactions that was still apparent six months after prosthodontic repair in one participant. Repeated baseline recordings and a control condition would help assess whether the repair accounted for the change.
3. From EEG Decoding to Real-World Performance
Offline accuracy tells us how well a decoder classifies recorded data. Performance during use also depends on how the system presents stimuli and responds to commands. Gangapuram and Manian [10] reported 96.6% accuracy on an offline motor imagery task using PhysioNet data. Their method classified sparse wavelet features without reconstructing the waveform, leaving an online test to establish whether skipping that step reduces response time. In a rapid visual presentation study with 24 healthy adults, Klee et al. [11] found that varying the intervals between stimuli made the steady-state visual evoked response less prominent in the averaged EEG.
The robotic arm study by Cardoso et al. [12] shows how much task time can depend on the interface itself. Eight participants without motor impairment used alpha modulation and an auditory menu to complete a pick-and-place task. Each task took about seven minutes on average, and much of that time was spent waiting for a menu cue. The arm was in motion for only about a tenth of the time. Vavoulis et al. [13] identified a related problem in their review of online motor imagery for stroke rehabilitation. Differences in participants, protocols, and feedback make it difficult to compare performance across systems.
4. Directions for Future Work
One priority for future work is to determine how much of an EEG result comes from the brain activity under study and how much comes from the way it was recorded and processed. Shared benchmarks could include repeat recordings from the same participants, with movement and muscle activity annotated. Researchers could then compare preprocessing pipelines on the same data and ask whether a waveform survives artifact removal or whether a connectivity finding persists when the reference changes. For brain-computer interfaces, pairing those recordings with logs of stimulus timing and user actions would help separate decoding errors from delays introduced by the interface. Such benchmarks would give researchers a way to explain why a method succeeds or fails under particular recording conditions.
Foundation models offer another direction for EEG analysis. EEG foundation models such as LaBraM and CBraMod learn representations across datasets and tasks [14,15]. Testing these models on participants and sites excluded from training could show how well they transfer across montages, recording environments, and patient populations. Comparisons with simpler methods designed for each task would help identify when pretraining improves transfer and when explicit signal processing remains useful. Reporting acquisition settings, preprocessing, and data splits would allow others to reproduce those comparisons.
We thank the authors, reviewers, and the Signals editorial team for their contributions to this Special Issue.
Conflicts of Interest
The author declares no conflict of interest.
References
- Ivaldi, M.; Giacometti, L.; Conversi, D. Quantitative Electroencephalography: Cortical Responses under Different Postural Conditions. Signals 2023, 4, 708–724. [Google Scholar] [CrossRef] [Scilit]
- Grobbelaar, M.; Phadikar, S.; Ghaderpour, E.; Struck, A.F.; Sinha, N.; Ghosh, R.; Ahmed, M.Z.I. A Survey on Denoising Techniques of Electroencephalogram Signals Using Wavelet Transform. Signals 2022, 3, 577–586. [Google Scholar] [CrossRef] [Scilit]
- Versaci, M.; La Foresta, F. EEG Data Analysis Techniques for Precision Removal and Enhanced Alzheimer’s Diagnosis: Focusing on Fuzzy and Intuitionistic Fuzzy Logic Techniques. Signals 2024, 5, 343–381. [Google Scholar] [CrossRef] [Scilit]
- Arrufat-Pié, E.; Estévez-Báez, M.; Estévez-Carreras, J.M.; Leisman, G.; Machado, C.; Beltrán-León, C. Beyond Frequency Band Constraints in EEG Analysis: The Role of the Mode Decomposition in Pushing the Boundaries. Signals 2023, 4, 489–506. [Google Scholar] [CrossRef] [Scilit]
- Abbasi, H.; Battin, M.R.; Butler, R.; Rowe, D.; Lear, B.A.; Gunn, A.J.; Bennet, L. Early Signatures of Brain Injury in the Preterm Neonatal EEG. Signals 2023, 4, 630–643. [Google Scholar] [CrossRef] [Scilit]
- Abbasi, H.; Battin, M.R.; Rowe, D.; Butler, R.; Gunn, A.J.; Bennet, L. CNN-Based Pattern Classifiers for Precise Identification of Perinatal EEG Biomarkers of Brain Injury in Preterm Neonates. Signals 2024, 5, 264–280. [Google Scholar] [CrossRef] [Scilit]
- Tozzi, A. Approaching Electroencephalographic Pathological Spikes in Terms of Solitons. Signals 2024, 5, 281–295. [Google Scholar] [CrossRef] [Scilit]
- Gangapuram, H.; Manian, V. Electroencephalogram Functional Connectivity Analysis and Classification of Mental Arithmetic Working Memory Task. Signals 2024, 5, 296–325. [Google Scholar] [CrossRef] [Scilit]
- Muroni, A.; Barbar, D.; Fraschini, M.; Monticone, M.; Defazio, G.; Marrosu, F. Case Report: Modulation of Effective Connectivity in Brain Networks after Prosthodontic Tooth Loss Repair. Signals 2022, 3, 550–558. [Google Scholar] [CrossRef] [Scilit]
- Gangapuram, H.; Manian, V. A Sparse Multiclass Motor Imagery EEG Classification Using 1D-ConvResNet. Signals 2023, 4, 235–250. [Google Scholar] [CrossRef] [Scilit]
- Klee, D.; Memmott, T.; Oken, B. The Effect of Jittered Stimulus Onset Interval on Electrophysiological Markers of Attention in a Brain–Computer Interface Rapid Serial Visual Presentation Paradigm. Signals 2024, 5, 18–39. [Google Scholar] [CrossRef] [Scilit] [PubMed]
- Cardoso, A.S.S.; Kæseler, R.L.; Jochumsen, M.; Andreasen Struijk, L.N.S. Manual 3D Control of an Assistive Robotic Manipulator Using Alpha Rhythms and an Auditory Menu: A Proof-of-Concept. Signals 2022, 3, 396–409. [Google Scholar] [CrossRef] [Scilit]
- Vavoulis, A.; Figueiredo, P.; Vourvopoulos, A. A Review of Online Classification Performance in Motor Imagery-Based Brain–Computer Interfaces for Stroke Neurorehabilitation. Signals 2023, 4, 73–86. [Google Scholar] [CrossRef] [Scilit]
- Jiang, W.-B.; Zhao, L.-M.; Lu, B.-L. Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI. arXiv 2024, arXiv:2405.18765. [Google Scholar]
- Wang, J.; Zhao, S.; Luo, Z.; Zhou, Y.; Jiang, H.; Li, S.; Li, T.; Pan, G. CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding. arXiv 2025, arXiv:2412.07236. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.