🔬 physics

Microwave focusing with temporal interference for non-invasive deep brain stimulation

This paper proposes and validates a non-invasive deep brain stimulation method that combines iterative time reversal and temporal interference optimization to precisely focus microwave-induced electric fields on specific brain targets within a realistic head model while ensuring safety through specific absorption rate compliance.

Original authors: Mika Söderström, Melker Carlsson, Patrik Nicolausson, Mariana Dalarsson

Published 2026-06-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Mika Söderström, Melker Carlsson, Patrik Nicolausson, Mariana Dalarsson

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a deep-sea treasure chest (a specific part of your brain) that needs a gentle nudge to work correctly. Currently, the only way to reach it is to drill a hole in the skull and insert a wire—a risky surgery. This paper proposes a new, non-invasive way to send that "nudge" using microwaves, like a high-tech flashlight that can shine only on the treasure chest without lighting up the rest of the room.

Here is how the authors achieved this, explained through simple analogies:

1. The Problem: The "Flashlight" Dilemma

Normally, if you try to shine a light (or an electric signal) through a thick, foggy wall (your skull and brain tissue), the light scatters and blurs. You can't focus it on a tiny spot deep inside without also blinding everything in front of it.

The Old Way: Low-frequency signals (like those used in current non-invasive treatments) are like a wide floodlight. They can pass through the skull, but they illuminate the whole brain, not just the specific spot you want.
The New Idea: Use high-frequency microwaves (like a laser) because they can be focused tightly. But there's a catch: brain cells (neurons) are too slow to react to a laser-fast signal. They only respond to slow, rhythmic pulses.

2. The Solution: The "Beat" Trick (Temporal Interference)

The authors use a clever trick called Temporal Interference (TI).

The Analogy: Imagine two people clapping their hands.
- Person A claps at a very fast, steady rhythm (a high-frequency wave).
- Person B claps at a slightly different, equally fast rhythm.
- If you stand in a spot where their claps perfectly overlap, you hear a "wah-wah-wah" sound (a beat) that is much slower than the individual claps.
How it works in the brain: The researchers send two high-speed microwave signals into the brain. Everywhere else, the signals are just fast noise that the brain ignores. But at the exact spot where the two signals meet, they create a "beat" (a slow, rhythmic envelope). This slow beat is the signal the brain cells can actually hear and respond to.

3. The Setup: The "Smart Speaker Array"

To make this happen, they didn't just use one antenna. They used an array of many small antennas surrounding the head, like a ring of smart speakers.

The Challenge: The brain is like a house made of different materials (bone, fat, water, gray matter). Some materials slow down the waves, some absorb them. It's a messy maze.
The Two-Step Strategy:
1. Step 1: The "Echo" Map (Iterative Time Reversal): Imagine shouting into a cave and listening to the echo. The computer simulates sending a signal from the target spot outward to the antennas, then reverses that path. This tells the system exactly where to place the antennas and how to aim them to fight through the messy brain tissue. It's like mapping the perfect path through a maze by walking it backward.
2. Step 2: The "Conductor" (Genetic Algorithm Optimization): Once the antennas are placed, the computer acts like a conductor for an orchestra. It tweaks the volume (amplitude), timing (phase), and which "note" (frequency) each antenna plays. It runs thousands of simulations to find the perfect combination that makes the "beat" loud and clear only at the target, while canceling it out everywhere else.

4. The Results: A Focused, Safe Beam

The authors tested this on a highly detailed, 3D digital model of a human head (with 38 different types of tissue).

Precision: They found they could focus the "beat" into a small, tight ball (about the size of a grape) deep inside the brain.
Safety: They checked the "heat" generated (Specific Absorption Rate or SAR). Just like a microwave oven heats food, these waves can heat tissue. The study showed that even with the high power needed to reach deep brain areas, the heat stayed well within safe limits, similar to what is allowed for standard medical devices.
Robustness: They tested what happens if the antennas are slightly moved or if the settings are slightly off. The system remained stable, meaning the "beam" didn't wobble dangerously.

5. The Trade-offs

The paper highlights a balancing act, like tuning a radio:

Frequency: Lower frequencies penetrate deeper but are harder to focus tightly (like a wide beam). Higher frequencies focus tightly but get absorbed by the skin and skull before reaching deep (like a laser that gets blocked by fog). The authors found a "sweet spot" (around 700 MHz) that balanced depth and focus.
Number of Antennas: More antennas allowed for a tighter, stronger focus. Fewer antennas resulted in a blurrier beam.

Summary

This paper presents a computer simulation proving that it is theoretically possible to use an external ring of antennas to send a focused, safe, and rhythmic "nudge" deep into the brain without surgery. They achieved this by combining a "backward-mapping" technique to navigate the brain's complex structure with a "beat-making" trick to make the brain cells react to high-speed waves.

Important Note: This is a computational study. The authors built a digital model and ran simulations. They did not test this on real humans or animals in this specific paper. They established the feasibility of the method in a realistic digital environment, setting the stage for future physical development.

Technical Summary: Microwave Focusing with Temporal Interference for Non-Invasive Deep Brain Stimulation

Problem Statement
Deep Brain Stimulation (DBS) is a proven treatment for neurological disorders such as Parkinson's disease, epilepsy, and depression, yet it currently relies on invasive surgical implantation of electrodes. This invasive approach carries risks including hemorrhage, infection, and hardware failure. While non-invasive alternatives like Transcranial Magnetic Stimulation (TMS) and Transcranial Electric Stimulation (TES) exist, they utilize low-frequency fields that lack the spatial selectivity required to target deep brain structures without affecting superficial tissue. Conversely, microwave (MW) fields offer improved spatial focusing potential but cannot directly stimulate neurons due to the neuron's low-pass filtering characteristics (cut-off frequencies typically below 1 kHz). The challenge is to achieve non-invasive, deep-brain targeting with high spatial precision while adhering to safety limits regarding tissue heating.

Methodology
The authors propose a computational framework that combines Iterative Time Reversal (iTR) and Temporal Interference (TI) optimization to generate focused, amplitude-modulated (AM) electric fields within a realistic human head model.

Simulation Environment: The study utilizes an anatomically realistic voxel head model (Duke from the IT'IS Virtual Population) comprising 38 unique tissue types with heterogeneous, lossy, and dispersive properties. The model is surrounded by a lossless matching layer and an external array of Magnetic Point Dipoles (MPDs) positioned on a Fibonacci sphere.
Two-Stage Optimization Framework:
- Stage 1: Iterative Time Reversal (iTR): This stage optimizes the physical configuration of the antenna array. A point source is placed at the desired focal point inside the head to generate a forward field. The system then reconstructs a backward field using time-reversed propagation. Through iterative updates of antenna positions, orientations, and steering parameters, the algorithm minimizes an objective function based on the Half-Intensity Radius (HIR) and Focus Intensity Quotient (FIQ). This provides a physically informed initialization for the array geometry.
- Stage 2: Temporal Interference (TI) Optimization: Using the array geometry fixed from Stage 1, a Genetic Algorithm (GA) optimizes the excitation parameters (amplitude, phase, and frequency assignment) for each antenna. Two high-frequency carriers ( $f_1$ and $f_2$ ) are used to generate an AM field with a low-frequency envelope ( $\Delta f$ ). The objective is to maximize the envelope amplitude ( $E_{AM}$ ) at the target while minimizing it elsewhere, effectively steering the stimulation to deep regions without activating superficial neurons.
Safety Evaluation: Compliance with international safety standards (IEEE/ICNIRP) is assessed using the Specific Absorption Rate (SAR), specifically the 10-g spatially averaged SAR ( $SAR_{10g}$ ), ensuring it remains below 10 W/kg.

Key Contributions

Unified Framework: This work presents the first computational demonstration combining iTR and TI optimization specifically for microwave-based non-invasive DBS in a realistic voxel head model.
Novel Objective Function: Unlike previous MW focusing studies that optimize single-frequency field intensity (often for hyperthermia), this method introduces a TI-based objective function that explicitly targets the neuromodulation mechanism (envelope amplitude).
Extension to High Frequencies: The framework extends TI principles from the traditional kHz carrier range to the MHz–GHz regime, enabling compatibility with microwave focusing capabilities.
Comprehensive Validation: The study includes systematic parameter sweeps (frequency, array size, target location), perturbation analysis, and statistical evaluation to demonstrate robustness against anatomical heterogeneity and implementation uncertainties.

Results
Numerical simulations across various configurations yielded the following findings:

Spatial Localization: The method successfully generated localized AM fields at target depths (e.g., subthalamic nucleus) with a focality volume ( $V_{50\%}$ ) corresponding to a radius of approximately 3–4 cm.
Intensity and Penetration: Peak envelope intensities ( $E_{max}^{AM}$ ) ranged from approximately 96 V/m to 178 V/m depending on frequency and target depth. Lower frequencies provided higher intensity but reduced focality, while higher frequencies improved focality at the cost of intensity due to tissue attenuation.
Safety: All configurations maintained $SAR_{10g}$ below the 10 W/kg limit in relevant tissues (skin, CSF, skull). The maximum SAR values were consistently observed in lateral superficial regions (near the ears) rather than the deep target.
Robustness: The system demonstrated stability against small perturbations in antenna position (±2 mm), orientation (±2°), magnitude (±1.5 $\mu$ Am $^2$ ), and phase. While the focal field remained stable, SAR values in superficial regions showed slightly higher sensitivity to perturbations, occasionally exceeding the limit by a small margin in worst-case scenarios.
Target Flexibility: The framework maintained consistent focusing performance for targets displaced by ±2 cm from the center, adapting antenna excitations to compensate for anatomical asymmetry and varying propagation paths.

Significance and Claims
The authors claim that this study establishes a foundational framework for safe, non-invasive microwave neuromodulation. By integrating iTR for array configuration and TI for stimulation mechanism, the approach addresses the dual challenges of deep penetration and spatial selectivity. The paper asserts that, to the authors' knowledge, this is the first study to apply this combined optimization strategy in a realistic head model for MW-DBS. The results suggest that external MW antenna arrays can achieve localized stimulation with intensities comparable to or higher than previous invasive TI-based MW studies, while avoiding the risks associated with implanted hardware. The work highlights the inherent trade-offs between focality, penetration depth, and safety, providing a basis for future development of physically realizable antenna systems and more detailed neurophysiological models.