Neurofeedback modulation of insula activity via MEG-based brain-machine interface: A double-blind randomized controlled crossover trial

Takufumi Yanagisawa, Yuhao Wang, Ryohei Fukuma, Ben Seymour, Huixiang Yang, Haruhiko Kishima

Introduction

The insular cortex is a critical hub in the brain’s "pain matrix," responsible for processing both the sensory intensity of pain and the emotional distress that accompanies it. In many chronic pain conditions, such as fibromyalgia, the insula is known to be hyperactive. While neuromodulation offers a potential therapeutic avenue, two fundamental questions remain: can humans learn to cognitively control this deep brain region, and does such control lead to a causal change in pain perception?

Existing neurofeedback technologies involve significant trade-offs. Electroencephalography (EEG) is fast but lacks the spatial resolution to accurately target deep structures like the insula. Functional MRI (fMRI) offers high spatial resolution but suffers from a multi-second delay in feedback, making real-time "brain training" difficult. This study explores whether Magnetoencephalography (MEG) can provide a "Goldilocks" solution, combining high spatial resolution with millisecond-level temporal precision for a real-time brain-machine interface (BMI).

Methods

Researchers conducted a double-blind, randomized controlled crossover trial with 19 healthy participants (11 males, 8 females; mean age 28.8). The study utilized an MEG-based BMI that employed Variational Bayesian multimodal encephalography (VBMEG) to estimate cortical currents at 225 specific vertices within the right insula.

Participants underwent training sessions for both the upmodulation and downmodulation of insula activity. During these sessions, they viewed a visual feedback stimulus—a black disk—that changed size based on the Root Mean Square (RMS) of the estimated currents in the target region. In upmodulation trials, the disk grew as activity increased; in downmodulation trials, it grew as activity decreased. Participants were tasked with discovering mental strategies to control the disk's size.

Thermal pain thresholds were measured before and after each training session using a "method of limits" applied to the participant's hand. To ensure high anatomical precision, individual T1-weighted MRI scans were used to create personalized 3D brain models, with the insula defined according to the Automated Anatomical Labeling (AAL) atlas.

Results

The study successfully demonstrated that humans can consciously regulate activity in the right insula. There was a highly significant statistical difference in insula activity between the upmodulation and downmodulation sessions ($p < 0.0001$). Notably, this modulation was observed in the overall magnitude of neural excitability (RMS) rather than in specific spectral power bands (such as Alpha or Theta), suggesting that the total volume of neural firing is a more effective target for pain-related neurofeedback.

Furthermore, the training appeared to induce short-term neuroplastic changes. Resting-state activity in the insula was significantly lower following downmodulation training compared to upmodulation training ($p = 0.003$).

Crucially, downmodulation training was associated with a significant increase in pain thresholds ($p = 0.006$), meaning participants could tolerate higher temperatures before reporting pain. While upmodulation did not significantly lower pain thresholds—possibly due to habituation to the thermal stimuli—the successful increase in tolerance following downmodulation provides a promising proof-of-concept for the causal role of the insula in pain modulation.

Discussion and Conclusion

These findings demonstrate that MEG-based neurofeedback can effectively target deep brain structures to modulate pain processing. The ability to "tune" insula activity through cognitive strategies represents a significant advancement in non-invasive neuromodulation.

However, several limitations warrant further investigation. The lack of a significant statistical interaction effect between the upmodulation and downmodulation groups suggests that some observed pain relief may be due to general placebo or habituation effects. Additionally, as this study was conducted on healthy volunteers, the next critical step is to test this system in clinical populations with chronic pain. Future research should also examine the "dose-response" of this training to determine if multiple sessions can lead to more durable or robust therapeutic outcomes. Despite these challenges, the study establishes a viable path for developing high-resolution, non-invasive MEG neurofeedback systems for clinical use.

Summary updated

Abstract

<title>Abstract</title> Insula activity has often been linked to pain perception, making it a potential target for therapeutic neuromodulation strategies such as neurofeedback. However, it is not known whether insula activity is under cognitive control and, if so, whether this activity is consequently causally related to pain. Here, we conducted a double-blind randomized controlled crossover trial to test the modulation of insula activity and pain thresholds using neurofeedback training. Nineteen healthy subjects underwent neurofeedback training for upmodulation and downmodulation of right insula activity using our magnetoencephalography (MEG) based brain–machine interface. We observed significant differences in insula activity between the upmodulation and downmodulation training sessions. Furthermore, resting-state insula activity significantly decreased following downmodulation training compared to following upmodulation training. Compared with upmodulation training, downmodulation training was also associated with increased pain thresholds, albeit with no significant interaction effect. These findings show that humans can cognitively modulate insula activity as a potential route to develop therapeutic MEG neurofeedback systems for clinical testing.
BibTeX
@inproceedings{yanagisawa2025neurofeedback,
  author = {Takufumi Yanagisawa and Yuhao Wang and Ryohei Fukuma and Ben Seymour and Huixiang Yang and Haruhiko Kishima},
  title = {{Neurofeedback modulation of insula activity via MEG-based brain-machine interface: A double-blind randomized controlled crossover trial}},
  booktitle = {Communications},
  year = {2025},
  doi = {10.1038/s42003-025-08176-8},
}
Figure