Dynamic Causal Modeling Advances TMS Therapy for Depression
Advanced computational techniques allow us to understand how the brain works, and suggests refined targets for interventional psychiatry with non-invasive brain stimulation.
Repetitive transcranial magnetic stimulation (rTMS) and other targeted, noninvasive neuromodulation approaches offer promise for treating neuropsychiatric conditions. These techniques work by stimulating key brain regions to increase or decrease activity in those areas. Results depend on which areas are stimulated and their interconnections with other regions—when we modulate activity in one spot, brain circuit connectivity means other areas may also be affected. TMS works through "cortical windows" to affect deeper brain areas, but our understanding of how this works is early.
The Complex, Nonlinear Brain
Emerging research on "dynamic causal modeling" (DCM) based on brain functional imaging (fMRI) will enable selective targeting to impact deeper brain structures indirectly. True deep noninvasive neuromodulation like transcranial focused ultrasound (tFUS) will allow us to reach subcortical structures without surgery, unlike deep brain stimulation (DBS), which, while effective, requires surgically implanted electrodes.
DCM uses probability-based inference—Bayesian epistemology—to build and test nonlinear models of complex systems, determining how system components and their connections lead to outcomes. For brain research, this means analyzing neural function and complex data sets to find causal levers affecting brain function. Causal modeling creates correspondence between what happens and what changes performance. This is currently lacking in psychiatric diagnostic systems, where there is scant causal understanding linking symptoms to treatment and outcome. DCM, coupled with artificial intelligence, provides a path toward better medical understanding of psychiatry.
Even "deep" TMS, despite being presented as reaching deeper brain structures, offers only marginally greater reach, actually being broader and affecting multiple brain regions nonselectively rather than much deeper. TMS can reach the brain's surface, but any impact on deep structures happens through brain interconnectivity, via the aforementioned cortical windows. While TMS is approved and well-studied clinically, other noninvasive approaches, including transcranial focused ultrasound stimulation (tFUS), which can reach deep structures directly, remain research-only.
This complexity stems from human brain organization and our nascent understanding of neuroscience. At a high level, we can understand neuromodulation as repairing brain networks, shifting them from dysfunctional states (dysconnectivity) to healthier, normalized states—what I call restoring "euconnectivity."
For example, if the central executive network (the frontoparietal network, or CEN) is relatively weak in depression compared with the default mode network, with insufficient capacity to regulate the salience network—affecting our ability to manage attention and regulate maladaptive emotional states—rTMS can correct this. Studies show that rTMS for major depressive disorder (MDD) significantly increases frontoparietal network activity.
Studying TMS Treatment of MDD Using Dynamic Causal Modeling
Groundbreaking work, elegant and beautiful, published in Nature's Translational Psychiatry (Kita et al., 2025), highlights DCM's power to understand how TMS affects brain systems in depression treatment, paving the way for personalized treatment design. Researchers compared 270 healthy controls with 175 people with MDD, analyzing resting-state fMRI to determine which brain areas and interconnections accounted for clinical response to rTMS.
The granular findings reveal several highlights. First, they confirmed that current rTMS approaches for treating major depression make causal sense. Multiple aberrant connections among diverse brain regions in depression-related circuits improved with left DLPFC (dorsolateral prefrontal cortex) stimulation. The thalamus (the brain's "switchboard" that routes information among areas) and visual cortex had altered functional connectivity with LDLPFC. Dysconnectivity was present across LDLPFC, the amygdala ("fear" center), nucleus accumbens (reward and dopaminergic neurotransmission), anterior insula (body perception and interoception, among other functions including empathy and motor control), subgenual anterior cingulate cortex (long associated with treatment-resistant depression, related to judgment and decision-making), and ventromedial prefrontal cortex (social cognition, emotion regulation, and decision-making).
In people with MDD, multiple connections charted a global roadmap of depression's neurocircuitry. DCM found decreased causal connections between DLPFC and visual cortex, between amygdala and anterior cingulate, amygdala and visual cortex, nucleus accumbens and visual cortex, anterior insula and nucleus accumbens, anterior insula and ventromedial prefrontal cortex, and decreased self-inhibition within the thalamus (suggesting lack of proper modulation in brain activity control). Depression severity linked with greater amygdala-anterior cingulate causal connectivity, and with lower causal connection between nucleus accumbens and visual cortex, and anterior insula and ventromedial prefrontal cortex.
This study demonstrates that clinical depression involves not just cortical brain activity, but clear connections with deep brain structures. While suspected, prior research hadn't clearly demonstrated the rich array of causal connectivity in MDD across the whole brain. DCM goes beyond conventional functional connectivity work to show the noted range of causal connections underlying depression. This means modifying these problems should improve depression aspects. Speculatively, visual cortex connections suggest people with depression literally (and of course, figuratively) see things differently—visual processing is altered. Amygdala and other emotional regulation region connections represent specific pathways in which "network surgery" could restore health. Body perception area connections, like the anterior insula, suggest avenues to address negative body image in depression and other conditions. Reward-based area connections like the nucleus accumbens account for impulse control difficulties and could be brain stimulation targets. Causal relationships with the anterior cingulate cortex (already targeted in "SAINT" TMS) suggest ways to improve decision-making and cognitive control of deeper brain structures.
Future Directions
In psychiatric conditions such as MDD and others—OCD, anxiety disorders, PTSD, ADHD, substance and alcohol use disorder, eating disorders—understanding the brain's rich, complex interconnectivity is imperative not just for understanding a specific condition, but for designing personalized treatment protocols for individualized care. While standard approaches such as left DLPFC treatment for MDD are quite effective, bespoke protocols may be necessary for different cases, such as depression subtypes or depressed patients with traumatic brain injuries. For OCD, with several different pathways already identified, we expect better results with individualized treatment planning based on brain imaging and augmented by artificial intelligence and machine learning, as brain connectivity mechanisms exceed average human comprehension. Unlike shoulder surgery, where surgeons can understand the complex mechanics of muscles, tendons, cartilage, and bone around the joint to make repairs, the brain is far too complex.
Future directions involve combining DCM with artificial intelligence to facilitate understanding of clinical depression and other psychiatric disorders. AI will likely vastly speed up, even automate, the process. We can envision a not-so-distant future in which interventional psychiatrists can order brain scans, have them analyzed to produce recommended treatment protocols, and use those protocols to alleviate problems by updating aberrant connections in the brain's operating system, and hence the mind. Furthermore, DCM and related work will enable better diagnostic systems based on actual brain function, rather than current systems based on clinical observation and statistical analysis of highly overlapping symptom clusters rather than empirical data.
For real-world brain stimulation use, such techniques are already available in early form. Our group uses TMS targeting based on brain connectivity research and will soon implement a neuronavigation system1, which, combined with individualized brain scans, will likely enable personalized, precision treatment. Understanding the role of combined treatments—for example, TMS and psychotherapy—will help us to not only better assist patients in finding relief from symptoms and by treating psychiatric disorders, but also to leverage neuroplastic effects of brain stimulation to positively bend the developmental curve, aspirationally altering the future course of a person's life for the better.
TMS is FDA on-label for the psychiatric treatment of clinical depression and OCD using standard approaches. Many of the techniques and methods presented here are off-label, and considering whether the possible benefit of off-label treatments with a sufficient evidence base is sufficient to offset the risks of receiving off-label treatment outside of a research setting should only be conducted with a qualified professional.
Kita, A., Ishida, T., Kita, N. et al. Exploring the capabilities of repetitive transcranial magnetic stimulation in major depressive disorder: Dynamic causal modeling of the neural network. Transl Psychiatry 15, 257 (2025). doi.org/10.1038/s41398-025-03480-7
TMS Outperforms Medication in Treatment-Resistant Depression
Developing More Effective, Accelerated TMS Protocols
Toward Brain-Based Psychiatric Diagnosis
Updating Psychiatric Diagnosis With Dynamic Causal Modeling
ExperiMentations Blog Post ("Our Blog Post") is not intended to be a substitute for professional advice. We will not be liable for any loss or damage caused by your reliance on information obtained through Our Blog Post. Please seek the advice of professionals, as appropriate, regarding the evaluation of any specific information, opinion, advice, or other content. We are not responsible and will not be held liable for third party comments on Our Blog Post. Any user comment on Our Blog Post that, in our sole discretion, restricts or inhibits any other user from using or enjoying Our Blog Post is prohibited and may be reported to Sussex Publisher/Psychology Today/Substack. Grant H. Brenner. All rights reserved.