Personalized Transcranial Electrical Stimulation: A Review of Computational Modeling and Optimization

TL;DR

This review articulates how inter-individual differences in brain anatomy and physiology impede uniform tES effects and reviews computational pipelines that enable personalized stimulation. It covers forward modeling (tissue segmentation, mesh generation, FEM-based field simulation, and software tools) and inverse optimization (leadfield-based formulations, constraints, and algorithms) for subject-specific montages. The authors highlight enhancements from multimodal imaging, anisotropy-aware conductivity, and neural dynamic coupling (neural mass models) to support state-aware, closed-loop tES. They identify limitations such as hardware variability, measurement artifacts, and the need for ground-truth data, advocating for integrated, data-driven, adaptive frameworks that move toward precision neuromodulation in both research and clinical contexts.

Abstract

Objective. Personalized transcranial electrical stimulation (tES) has gained growing attention due to the substantial inter-individual variability in brain anatomy and physiology. While previous reviews have discussed the physiological mechanisms and clinical applications of tES, there remains a critical gap in up-to-date syntheses focused on the computational modeling frameworks that enable individualized stimulation optimization. Approach. This review presents a comprehensive overview of recent advances in computational techniques supporting personalized tES. We systematically examine developments in forward modeling for simulating individualized electric fields, as well as inverse modeling approaches for optimizing stimulation parameters. We critically evaluate progress in head modeling pipelines, optimization algorithms, and the integration of multimodal brain data. Main results. Recent advances have substantially accelerated the construction of subject-specific head conductor models and expanded the landscape of optimization methods, including multi-objective optimization and brain network-informed optimization. These advances allow for dynamic and individualized stimulation planning, moving beyond empirical trial-and-error approaches.Significance. By integrating the latest developments in computational modeling for personalized tES, this review highlights current challenges, emerging opportunities, and future directions for achieving precision neuromodulation in both research and clinical contexts.

Figures (3)

• Figure 1: Overview of the forward modeling and inverse optimization for personalized tES. (A) The construction of forward models for personalized computational simulations begins with tissue segmentation from MRI data. (B) Conductivity values are assigned to each segmented tissue type to generate an individualized computational model. (C) Head conductor model is generated by mesh generation and numerical methods. (D) Virtual electrodes are then positioned on the model, and current is applied according to the predefined stimulation protocol. (E) The electric field distribution is simulated by solving the current flow within the head model. The stimulation montage is optimized based on the electric field simulation results from the forward model. (F) The stimulation target is first identified, followed by (G) verifying current safety constraints. (H) An optimization algorithm is then applied to the model to determine optimal stimulation parameters. (I) Based on the optimization results, a stimulation montage is applied to the personalized tES for the subject. Steps (F–I) illustrate how the stimulation target, safety constraints, and optimization algorithms can use forward model results to determine optimal stimulation parameters. In this section, the focus remains on forward modeling, while inverse optimization is briefly illustrated to provide context on the overall computational workflow.
• Figure 2: Schematic diagram of the stimulation optimization methods. (A) Reciprocity principle: The Reciprocity Principle posits that the path of electrical current flow from a stimulation electrode placed on the scalp to a designated area within the brain is equivalent to the path of current flow in the reverse direction, from that specific brain area back to the scalp electrode. Consequently, this principle allows for the utilization of a dipole and a simulation model to determine the optimal electrode configuration for targeted stimulation. (B) Convex optimization programming: By iteratively substituting a non-convex problem with solvable convex relaxations or linearized surrogates and refining them, one can leverage convex optimizers to efficiently approach the global optimum of the original formulation. (C) Genetic algorithm: This algorithm encodes electrode configurations into the genes of each solution within a population. Genes represent the activation status of the corresponding channel. The solutions represent a stimulation pattern.During each iteration, the solutions undergo crossover and mutation processes at random, resulting in the generation of offspring. Offspring demonstrating superior performance are retained. Ultimately, we can get the optimal solution at the last iteration. (D) Deep learning: The approach utilizes a neural network in which a constant unit replaces the input. Fully connected layers are used to generate weights within the electrode layer, which features twice as many nodes as the number of electrodes. Utilizing both the electrode layer and the lead field matrix, the distribution of the electric field and then the loss is calculated. The goal is to optimize the electrode currents that minimize the loss by training the weights of the electrode layer. The panel is adapted from bahn2023computational.
• Figure 3: The influence of model updates on the norm of the electric field. (A) The Standard Model: The upper panel displays a segmentation with five tissue types, each assigned conductivities based on standard values from prior research (WM: 0.126; GM: 0.275; CSF: 1.654; Bone: 0.01; Scalp: 0.465). The lower panel illustrates the distribution of the electric field's magnitude under a 1 mA tDCS from F3 to F4 in the EEG10-10 system. Following visualizations follow the same settings. (B) Bone subdivisions. The bone is segmented into two distinct categories: compact bone with a conductivity of 0.008 and spongy bone with a conductivity of 0.025. (C) Conductivity optimization. The conductivity of Gray Matter is adjusted to its highest reported value in previous research, being 0.6. In practice, conductivity optimization can be combined with sEEG to fine-tune the model's parameters. (D) Anisotropy model. The upper panel depicts the conductivity of White Matter. The conductivity of tissues is configured as anisotropic by direct mapping based on linear rescaling of tensors from diffusion MRI. The upper threshold for conductivity is established at 2 S/m, with a maximum ratio of 10 between the largest and smallest conductivity eigenvalues. The unit of conductivity: S/m. The unit of electric field: V/m. Abbr: WM, White Matter; GM, Gray Matter; CSF, Cerebrospinal Fluid; EF, electric field; tDCS, transcranial Direct Current Stimulation.