Optimization of Multipolar Temporal Interference Stimulation in Personalized Head Models
DOI: Pending DOI
Synopsis
Temporal Interference (TI) stimulation is a modality of Transcranial Electrical Stimulation that uses the interference of two high frequency electric fields with slightly different frequencies. While being non-invasive, TI has the potential of reaching deep brain targets without exciting the overlying neural tissue. However, the nature of quasi-static fields and the tissue inhomogeneity render the focusing of TI in subcortical structures difficult. A multipolar extension of TI (mTI) was recently proposed in the literature to overcome these difficulties. In this work we present an optimization of mTI based on synthesis of standard TI patterns, and demonstrate with the help of personalized head models the potential of mTI for reaching and focusing in deep subcortical targets.
Background
Temporal Interference stimulation is a relatively recent brain stimulation modality with the potential of reaching deep brain targets non-invasively. It is a variation of transcranial alternating current stimulation (tACS), where two electrical signals at slightly different kHz frequencies are applied at two different electrode pairs respectively. The basic assumption behind TI is that neurons react mostly to the envelope of the total signal rather than to its high frequency components. For suitable electrode configurations the main locus of interference can be subcortical, thus affecting primarily deeper neural tissue. The ability of TI to stimulate deep brain targets in mice was first demonstrated by Grossman et al.(1).
Despite successful demonstrations(1,2,3,4,5,6) of the ability of TI to stimulate deep targets in humans, the achievable focality and targeting accuracy has its limits. This is why recently a multipolar extension of TI (mTI) was proposed(7,8), where more than one pair of frequencies is used to shrink the main stimulation region. The increased focality of mTI has been experimentally demonstrated(7), but a systematic optimization for mTI has been lacking so far. This work attempts to fill this gap with an optimization scheme for targeting and focusing mTI in subcortical targets. With the help of in-silico personalized models a pipeline was constructed that can simulate the TI intensity in the brain. By optimizing combinations of standard TI patterns it is demonstrated that mTI can improve the targeting and focusing of stimulation in deep subcortical targets.
Methods
This study was done entirely in-silico with a simulation and optimization pipeline which includes both readily available tools and new algorithms that were developed for this work.
The standard TI configuration consists of two pairs of electrodes on the scalp of the subject. Each pair is fed by a harmonic current with its own frequency in the kHz range (1kHz-10kHz). The difference of the two frequencies is below 100 Hz. In order to calculate the electric field produced by a pair of electrodes we used the open source SimNIBS software(9). The dataset that was used for the simulation (T1 and T2 structural MRI) is the “ernie” example subject provided by SimNIBS itself. Given the electric fields corresponding to two electrode pairs, a novel averaging technique was used to calculate the mean TI modulation depth.
Next we sampled the space of all possible TI electrode configurations by uniformly sampling the allowed placement positions on the example subject’s scalp with Ne=188 samples. This resulted in ca. 154 million possible TI configurations. Through unbiased random sampling of these configurations we calculated which brain voxels can be stimulated with TI patterns under various focality constraints. This gave rise to reachability maps of the brain, where each voxel stores the possible TI configurations that stimulate it.
Finally, for a given sub-cortical target, the standard TI combinations that reach any of its voxels where found, and among these combinations the optimal pair of TI patterns was selected which combined yield the mTI pattern with the best focality.
Results
This main results of this study concern:
· the ability of standard TI to stimulate subcortical targets and its limitations,
· the effectiveness of combining multiple TI patterns into mTI patterns for increasing the stimulation focality.
In order to quantify the ability of standard TI to stimulate deep brain structures, we performed a systematic parameter sweep of the possible electrode locations by uniformly sampling the scalp of the example subject. We generated 188 possible electrode positions which resulted in 154 484 253 possible TI configurations. Out of them an unbiased random sample of 624 272 TI combinations was selected and the respective stimulation patterns were simulated. This allowed us to generate a reachability map, where a voxel is marked as reachable only if it belongs to a monofocal TI pattern of total volume <1cm3. The percentage of excitable brain tissue (gray and white matter) that can be stimulated monofocally with sub-cm resolution was found to be at least 20%. In detail:
· > 16.00 % of gray matter
· > 24.99 % of white matter
can be stimulated monofocally with sub-cm resolution.
However, when we computed the percentage of gray matter that can be reached by TI under the same focality constraints for some typical subcortical regions, the results were much lower. In particular, only
· 0.26 % of hippocampus
· 0.95 % of amydgala
· 0.53 % of thalamus
can be stimulated monofocally with sub-cm resolution.
The above results show the limitations of standard TI for sub-centimeter focusing in subcortical structures. One possible explanation for this difficulty is the fact that TI patterns in the brain have high spatial variability, whose main cause is the tissue inhomogeneity. In the specific case of subcortical structures it is often the case that they are surrounded by white matter of lower conductivity than gray matter, which causes a local drop in the electric field and makes it difficult to focus the stimulation in their interior.
As a next step we investigated how the focality in deep structures can be improved with multipolar TI configuration. Each mTI pattern was synthesized by combining two standard TI patterns. The basis for selecting the latter was another reachability map for standard TI, where the focality restrictions were relaxed: each voxel was marked reachable if it belonged to a TI pattern consisting of at most 3 voxels clusters of total volume <10cm3. To synthesize a mTI pattern we selected the standard TI patterns with maximum overlap at the target and minimum overlap elsewhere. This algorithm was tested for the left hippocampus and left thalamus.
In the attached figure the increase of focality by superposition of TI patterns is demonstrated for a specific case in the hippocampus. In the left upper and lower panel two different TI patterns are shown, whose mTI superposition is shown in the middle panel. A zoom into the focal region of mTI is shown in the upper left panel. In the lower left panel a histogram demonstrates how much the focal volume of mTI decreases in relation to the focal volumes of the two TI patterns.
For the left hippocampus we found 69 standard TI patterns that overlap at the target. Choosing only the TI pairs with maximum overlap left us with 156 possible mTI combinations. Performing a statistical analysis over all of these combinations we found that:
· Mean / minimum focal volume of mTI: 20.35 mm3 / 5 mm3
· Mean / minimum focal volume of standard TI: 62.03 mm3 / 12mm3
· Maximum focality improvement of mTI vs TI: 88%
For the left thalamus 69 standard TI patterns and 156 mTI optimal combinations were found with the following statistics:
· Mean / minimum focal volume of mTI: 100.8 mm3 / 9 mm3
· Mean / minimum focal volume of standard TI: 125.8 mm3 / 17 mm3
· Maximum focality improvement of mTI vs TI: 94%
Discussion
Our in-silico analysis demonstrates both the potential and the limitations of standard TI stimulation, as well as ways to overcome said limitations with mTI configurations.
Our parametric study of standard TI showed that this modality is able to stimulate focally with sub-centimeter resolution a non-negligible part of the brain (>20% of combined gray and white matter). However, it is difficult or impossible to reach many deep subcortical targets with the standard configuration of two electrode pairs and two frequencies. One important reason for this limitation, which is highlighted by our simulations, is the inhomogeneous tissue structure in the vicinity of subcortical structures.
Furthermore, we showed that multipolar TI (mTI) is able to improve the focality of standard TI for certain subcortical targets. However, a statistical analysis of our results showed that the synthesis of mTI patterns from standard TI patterns does not always yield better results, hence the stimulation parameters must be determined using carefully designed optimization procedures in order to reap the benefits of the increased degrees of freedom of mTI. This optimization in turn relies heavily on the spatial structure of the underlying electric fields, which makes personalization a necessary step to achieve satisfactory results.
Given the large variety of neurological and psychiatric disorders that stimulation of deep subcortical targets can address(10), the ability to reach said targets non-invasively opens new avenues for clinical applications of TI and mTI stimulation modalities.
Disclosures
With financial support from Agence Nationale de la Recherche under France 2030 bearing the reference ANR-24-RRII-0005, on funds administered by Inserm.
References
1. Grossman, N. et al. Noninvasive Deep Brain Stimulation via Temporally Interfering Electric Fields. Cell 169, 1029-1041.e16 (2017).
2. Lee, S., Lee, C., Park, J. & Im, C.-H. Individually customized transcranial temporal interference stimulation for focused modulation of deep brain structures: a simulation study with different head models. Sci Rep 10, 11730 (2020).
3. Bahn, S., Lee, C. & Kang, B. A computational study on the optimization of transcranial temporal interfering stimulation with high‐definition electrodes using unsupervised neural networks. Human Brain Mapping 44, 1829–1845 (2023).
4. Violante, I. R. et al. Non-invasive temporal interference electrical stimulation of the human hippocampus. Nat Neurosci 26, 1994–2004 (2023).
5. Demchenko, I. et al. Target engagement of the subgenual anterior cingulate cortex with transcranial temporal interference stimulation in major depressive disorder: a protocol for a randomized sham-controlled trial. Frontiers in Neuroscience 18, 1390250 (2024).
6. Weise, K. et al. A leadfield-free optimization framework for transcranially applied electric currents. Computers in Biology and Medicine 195, 110648 (2025).
7. Botzanowski, B. et al. Focal control of non-invasive deep brain stimulation using multipolar temporal interference. Bioelectron Med 11, 7 (2025).
8. Williamson, A. et al. Enhanced temporal interference stimulation: pulsed and multipolar techniques for improved deep brain modulation. Brain Stimulation 18, 220–221 (2025).
9. Thielscher, A., Antunes, A. & Saturnino, G. B. Field modeling for transcranial magnetic stimulation: A useful tool to understand the physiological effects of TMS? in 2015 37th annual international conference of the IEEE (EMBC) 222–225 (IEEE, 2015).
10. Lozano, A. M. et al. Deep brain stimulation: current challenges and future directions. Nature Reviews Neurology 15, 148–160 (2019).