Transcranial Electrical Stimulation (TES)

• Transcranial Electrical Stimulation is a group of non-invasive techniques that deliver weak currents through scalp electrodes to modulate neural activity.
• Advanced electrode montage design and computational modeling optimize the spatial distribution of electric fields for safe and effective stimulation.
• TES modalities, including tDCS, tACS, and tTIS, are applied in cognitive enhancement, neuropsychiatric treatments, and targeted deep brain modulation.

Transcranial Electrical Stimulation (TES) is a family of non-invasive neuromodulation techniques in which weak currents are delivered through scalp electrodes to modulate neural activity in targeted brain regions. TES spans several core modalities—including transcranial direct current stimulation (tDCS), alternating current stimulation (tACS), random-noise stimulation (tRNS), pulsed current stimulation (tPCS), temporal interference stimulation (tTIS), and specialized research paradigms—each with specific waveforms, mechanisms, and clinical applications. Rigorous modeling, optimization, and experimental validation have positioned TES as an essential tool for both basic neuroscience and applied clinical practice.

1. Physical Principles and Core Modalities

Under the quasi-static approximation, electric fields in TES are governed by the Laplace equation $\nabla \cdot (\sigma(x) \nabla \phi(x)) = 0$, where $\sigma(x)$ is tissue conductivity and $\phi(x)$ is the scalar potential. Injection of weak currents (typically $0.5$–$4$ mA) via scalp-mounted electrodes induces cortical fields on the order of $0.2$–$1$ V/m, sufficient to modulate neuronal membrane potentials and influence population-level oscillatory activity (Tarazona et al., 2015).

TES encompasses several principal stimulation types:

• tDCS: Constant amplitude DC current; polarity-dependent modulation (anodal depolarization, cathodal hyperpolarization); durations of 10–30 min; interventional fields for rehabilitation and neuropsychiatric disorders (Cukic, 2019).
• tACS: Sinusoidal currents (0.5–2 mA, 1–140 Hz) for entrainment and phase modulation; used in studies of oscillatory synchrony and cognitive enhancement (Modolo et al., 8 Oct 2024, Liu et al., 2020).
• tRNS: Stochastic current waveforms for broad-band excitability modulation.
• tPCS/CES/MET: Pulsed, burst, or microcurrent variants with specialized digital controllers for clinical psychiatry/neurology (Jafari et al., 2020).
• tTIS: Two high-frequency channels (kHz-range), generating a low-frequency modulation envelope that uniquely targets deep brain structures such as hippocampus or striatum (Vassiliadis et al., 16 Dec 2025, Yatsuda et al., 14 Nov 2025).

2. Electrode Montage, Field Shaping, and Biophysical Constraints

TES efficacy fundamentally depends on the spatial distribution of electric fields, which are steered by electrode geometry, tissue conductivity, and optimization algorithms:

• Montage Design: Conventional setups use two large saline-soaked pads in bipolar montages, but high-definition arrays (4×1 ring, concentric, or multi-site patterns) enable increased focality (Saturnino et al., 2019). For instance, a concentric-ring montage with a central anodal cluster and outer cathodal ring (≥3 cm spacing) efficiently channels current through a cortical ROI (e.g., angular gyrus) (Pastor et al., 23 Dec 2025).
• Safety Constraints: Per-electrode currents are typically capped at $1$–$2$ mA, total current at $2$–$4$ mA, and current density at $0.1$ mA/cm² to prevent skin lesions (Tarazona et al., 2015).
• Field Optimization: Quadratic and linear programming, as well as modern evolutionary multi-objective algorithms (e.g., MOVEA (Wang et al., 2022), HingePlace (Goswami et al., 3 Feb 2025)), allow rigorous trade-offs among focality, intensity, depth, and avoidance zones, often via sparse multi-channel montages (optimal $k=6$–$8$ electrodes) (Saturnino et al., 2019).
• Tissue Conductivity: Inter-individual anatomical variation (scalp, skull, CSF, GM, WM) critically alters field distribution; recommended ranges based on in-vivo estimates: scalp 0.33–1.0 S/m, skull 0.0042–0.05 S/m, CSF ~1.79 S/m, gray matter 0.33–1.0 S/m, white matter 0.14–0.48 S/m (Altakroury et al., 2022).
• Segmentation and Model Accuracy: Segmentation uncertainties (Dice coefficient) in CSF and GM introduce 4–6% errors in peak field estimation; prioritizing high accuracy for these compartments significantly improves reliability (Rashed et al., 2020).

3. Mechanisms of Action and Neural Modulation

TES modulates neuronal populations through subthreshold membrane polarization, spike-timing bias, network-level phase locking, and ultimately plasticity:

• tDCS Mechanisms: Steady fields (0.2–0.8 V/m) induce slow shifts in resting membrane potential ($$\Delta V_m \approx -E \cdot \lambda \cdot \cos \theta$$), facilitating or inhibiting neuronal firing depending on polarity (Tarazona et al., 2015, Cukic, 2019).
• tACS Mechanisms: Frequency-specific sinusoidal fields entrain ongoing oscillatory networks, yielding "Arnold tongues" in amplitude–frequency space where phase-locking is maximal. Plastic aftereffects, lasting up to an hour, are observed via spike-timing-dependent plasticity (Modolo et al., 8 Oct 2024, Liu et al., 2020).
• tTIS Mechanisms: Two kHz-range AC currents produce a low-frequency envelope, which is rectified by voltage-gated sodium channels, leading to focal modulation in deep structures without direct surface activation. Clinical and animal studies confirm selective modulation of striatum and hippocampus (Vassiliadis et al., 16 Dec 2025, Yatsuda et al., 14 Nov 2025).

Complexity metrics (e.g., Higuchi's fractal dimension, sample entropy) in resting-state EEG are sensitive to TES-induced changes and show promise as biomarkers for clinical efficacy, particularly in depression (Cukic, 2019).

4. Computational Modeling, Forward/Inverse Solutions, and Data-Driven Pipelines

Modeling, optimization, and real-time emulation are essential in translating TES from bench to bedside:

• Forward Modeling: Personalized finite-element (FEM) or boundary-element (BEM) head models simulate field distributions given electrode layout and tissue properties; pipelines include ROAST, SimNIBS, and Zeffiro Interface (Wang et al., 1 Sep 2025, Prieto et al., 2022).
• Inverse Optimization: Designing montages to maximize targeting and minimize off-target effects leverages quadratic/linear programs (reciprocity, focality-constrained LP), genetic algorithms, and deep learning surrogates. Multi-objective frameworks provide Pareto-optimal solution fronts (Wang et al., 2022, Goswami et al., 3 Feb 2025, Prieto et al., 2022).
• Fast Emulation: Attention U-net architectures (e.g. DeeptDCS) predict 3D current density distributions in <1 s per sample, enabling rapid uncertainty quantification, grid-based montage search, and closed-loop adaptive TES (Jia et al., 2022).
• Boundary Electrode Modeling: CEM, GAP, and PEM approaches vary in their representation of skin–electrode interface impedance and heating; GAP models are generally sufficient for brain-field estimation, with CEM reserved for detailed safety or heating analysis (Agsten et al., 2016).
• Conductivity Calibration: In-vivo estimation using simultaneous EEG/sEEG with intracerebral stimulation refines subject-specific conductivities and reduces modeling uncertainty (Altakroury et al., 2022).

5. Applications in Basic and Clinical Neuroscience

TES modalities contribute to a broad spectrum of research and therapy:

• Consciousness Measurement: Reliable, region-targeted TES (e.g., posterior tDCS to angular gyrus) combined with deep learning EEG classification frameworks achieves F1-scores up to 92%—substantially better than human-level accuracy for evoked-state discrimination (Pastor et al., 23 Dec 2025).
• Cognitive Enhancement: tDCS and tACS show robust short-term improvements in working memory, language comprehension, and motor control (effect sizes 0.3–0.5) (Tarazona et al., 2015, Mastakouri et al., 2017).
• Treatment of Neuropsychiatric Disorders: tDCS has documented efficacy in resistant depression (complexity decrease as a biomarker), schizophrenia (oscillatory entrainment), epilepsy (network desynchronization), and chronic pain (Cukic, 2019, Schütt et al., 2012).
• Deep Brain Neuromodulation: tTIS enables non-invasive targeting of hippocampus and striatum, opening applications in Parkinson’s and Alzheimer’s disease and rehabilitation after stroke (Vassiliadis et al., 16 Dec 2025).
• Personalized/Tailored Protocols: EEG-based decoding models predict individual TES parameter sets for optimal motor rehabilitation and cognitive outcomes (Mastakouri et al., 2017).
• Digital Stimulator Engineering: Stand-alone programmable TES units support broad waveform generation (tDCS, tPCS, CES, MET) with high precision and safety monitoring for research and clinical environments (Jafari et al., 2020).

6. Limitations, Controversies, and Future Directions

TES is limited principally by field focality, model uncertainty, and inter-individual anatomical variability:

• Focality–Depth Trade-Off: Sufficiently focal and deep cortical targeting remains challenging due to biophysical constraints (skull resistivity, tissue anisotropy); multi-channel montages and advanced optimization algorithms provide partial remedies but not millimetric precision (Saturnino et al., 2019, Yatsuda et al., 14 Nov 2025, Vassiliadis et al., 16 Dec 2025).
• Modeling Uncertainties: Variability in segmentation accuracy and tissue conductivity propagates to field estimates; best practice requires reporting and propagating uncertainty (Rashed et al., 2020, Altakroury et al., 2022).
• Mechanistic Controversies: Peripheral nerve stimulation and skin shunting complicate interpretation of tACS/tDCS effects; rigorous controls and multimodal validation (e.g., tAMS, phase-linear controls) are required (Modolo et al., 8 Oct 2024).
• Closed-Loop and Adaptive TES: Active research is directed toward EEG/MEG-informed real-time adjustment of stimulation parameters (intensity, frequency, spatial targeting), optimizing responsiveness to individual brain state and network dynamics (Wang et al., 1 Sep 2025, Pastor et al., 23 Dec 2025).
• Deep Structure Targeting: Integration of tractography-informed optimization and epicranial electrode approaches enhances deep-network selectivity but requires further validation (Yatsuda et al., 14 Nov 2025).

Table: Representative TES Modalities, Waveforms, and Target Regions

Modality	Waveform/Parameters	Anatomical Target(s)
tDCS	DC, 1–2 mA, 10–30 min	Prefrontal, motor, angular gyrus
tACS	AC, 5–140 Hz, 1–2 mA	Visual, motor, working memory
tTIS	2×kHz AC, Δf=10–80 Hz	Striatum, hippocampus
tPCS	Pulsed, 0.5–1000 Hz	Clinical psychiatric targets
HD-Array/Montage	Multi-site, 6–32 electrodes	Focal cortical, deep subcortical
Personalized TES	Data-driven, feature-based	Individualized cortical network

7. Open Resources, Data, and Code Availability

Recent studies have prioritized reproducibility and open science:

• EEG+TES Datasets: Comprehensive datasets, preprocessing scripts, and trained models for deep learning-based consciousness measurement (Pastor et al., 23 Dec 2025):
  • Code: https://github.com/alexispomares/DL-EEG-TES
  • Raw Data: https://www.kaggle.com/alexispomares/dissertation-raw
  • Preprocessed: https://www.kaggle.com/alexispomares/dissertation-preprocessed
• Optimization Frameworks: MOVEA multi-objective codebase: https://github.com/ncclabsustech/MOVEA.
• Digital Stimulator Hardware: Schematics and firmware for multi-modal digital TES units (Jafari et al., 2020).
• Emulation Models: DeeptDCS: rapid field estimation (Jia et al., 2022).

TES research leverages cross-disciplinary advances in computational modeling, deep learning, hardware engineering, and clinical neuroscience. The trajectory points toward precision neuromodulation, individualized dosing, and integration with real-time brain-state monitoring. Rigorous adherence to modeling best practices, empirical validation, and open sharing of code and data remain central to future progress.