Imagine trying to repair faulty wiring in a house. You could rely on a standard plan, or first check where the wires run in that particular building. Transcranial magnetic stimulation faces a similar challenge.
Transcranial magnetic stimulation (TMS) uses magnetic pulses from a coil placed on the scalp to influence brain activity. It is an established evidence-based treatment for depression, although (like any treatment) clinical response varies between individuals (NICE, 2015). Evidence in post-traumatic stress disorder is less mature and more inconsistent (Brown et al., 2024).
Treatment is often described as “targeted”. However, clinicians commonly locate the stimulation site using scalp measurements or standard brain coordinates (Sack et al., 2009), assuming that the same location will reach the relevant brain circuit in everyone. But, brain networks vary between individuals.
Functional magnetic resonance imaging (fMRI) may offer a more personalised approach (Cash & Zalesky, 2024; Sack et al., 2009). It can identify an accessible brain area connected to deeper regions thought to be involved in a person’s symptoms (Cash et al., 2021; Cole et al., 2022). However, prospective trials are needed to determine whether individualised targeting improves clinical outcomes (Cash & Zalesky, 2024). Two recent trials by Taylor et al. (2026) and van Rooij et al. (2026) tested whether this more personalised approach could improve treatment, asking:
Can brain imaging help choose a better target, and does that translate into better outcomes for patients?
Methods
Three questions were asked:
- Where to stimulate? (target selection)
- Does the coil accurately reach that point? (neuronavigation)
- Should the target differ between individuals? (personalisation)
Taylor et al. (2026) randomised 40 adults with treatment-resistant depression to accelerated stimulation (ten sessions daily for five days) in a triple-blind design. The target was selected using either the site most strongly connected to a previously defined depression circuit on each person’s resting-state scan, or the Beam F3 scalp-measurement method (which estimates the left prefrontal site from head measurements). Both groups received neuronavigated stimulation. The primary outcome was clinician-rated depression severity one month after treatment.
Van Rooij et al. (2026) randomised 50 adults with post-traumatic stress disorder symptoms to active or sham low-frequency stimulation, delivered twice daily for ten days. Brain imaging identified the point in the right dorsolateral prefrontal cortex most strongly connected to the right amygdala, a brain region involved in processing threat. The primary outcomes were amygdala responses to threat and changes in sweating during trauma recall; symptom measures were secondary outcomes.
Results
Did connectivity-based target selection improve depression?
Both groups experienced substantial reductions in depression, and the trial produced a promising estimate of additional benefit from connectivity-based targeting. One month after treatment:
- the median reduction on the Montgomery-Åsberg Depression Rating Scale was 24 points with connectivity-based targeting
- the median reduction was 18 points with scalp-based targeting
- the difference was statistically significant (p = 0.02)
- the estimated effect size was 0.8, with a 95% confidence interval from 0.26 to 1.54 (Taylor et al., 2026).
However, the advantage was not apparent immediately after treatment or one week later, nor on a self-rated depression measure.
Clinical response, defined as a symptom reduction of at least 50%, occurred in 16 of 20 people (80%) receiving connectivity-based targeting and 12 of 20 (60%) receiving scalp-based targeting. From this 20 percentage point difference, the authors calculated a “number needed to scan” of five for one additional response. With only 20 people in each group, this is a hypothesis-generating estimate rather than a robust figure for clinical decisions.
For personalisation to make sense, targets should be reasonably stable within a person and genuinely differ between people; otherwise, scanner noise could be mistaken for individuality. Targets calculated from separate halves of the same participant’s scan were a median of 4.47 mm apart, compared with 12.97 mm between different participants (Taylor et al., 2026). This is consistent with the idea that targets were more similar within individuals than between individuals, although it was achieved with a functional scan that took 41 minutes, which could be a practical barrier to using personalised targeting routinely.
Did stimulating a personalised target help post-traumatic stress disorder?
In the second trial, active stimulation reduced right-amygdala responses to threat more than sham treatment (p = 0.035). However, this result did not survive the authors’ correction for multiple comparisons, although it did in a women-only subgroup analysis. This suggests that stimulation may have had an effect on one of its intended biological targets, although a change in brain response does not by itself show that it caused clinical improvement. The groups did not differ significantly on the other primary outcome: sweating responses during trauma recall, which provided a physiological measure of the body’s response to remembering the trauma (van Rooij et al., 2026).
Immediately after treatment, hyperarousal and overall post-traumatic stress disorder symptoms improved in both groups, with no significant differences between them. At the 3 to 6-month follow-up, active stimulation was associated with greater reductions in hyperarousal (p = 0.018) and total symptoms (p = 0.010) than sham stimulation (van Rooij et al., 2026).
This delayed clinical separation is interesting, but it requires caution. The post-treatment questionnaire asked about the preceding week, partly overlapping with treatment, which the authors suggest may have diluted an early effect. Conversely, only 34 of 47 participants contributed follow-up data, and 17 received open-label active stimulation beforehand. The authors’ sensitivity analyses are reassuring, but the later effect is less cleanly attributable to the original randomisation.
Conclusions
Taylor et al. (2026) concluded that individual connectivity-based target selection enhanced the antidepressant effects of accelerated stimulation compared with scalp-based targeting. Van Rooij et al. (2026) concluded that personalised imaging-guided stimulation reduced amygdala threat responses and improved longer-term post-traumatic stress disorder symptoms compared with sham treatment.
Both studies support further research into circuit-guided treatment, but they should not be treated as equivalent evidence.
- The depression trial tested whether imaging could select a better target: the results suggest it might, but this was a small 40-person pilot study.
- The post-traumatic stress disorder trial tested whether active stimulation at a personalised target performed better than sham treatment: there was evidence of an effect on one of two neurobiological outcomes, alongside a delayed clinical signal that is less certain because of missing follow-up data and subsequent open-label treatment.
Strengths and limitations
Both trials asked focused questions and used random allocation. Blinding of participants and assessors limited the influence of expectations on the results. The sham condition in the post-traumatic stress disorder trial also helped separate the effects of active stimulation from attention, repeated assessment and natural symptom change.
The depression trial used a useful head-to-head comparison. Both groups received the same protocol, making the targeting method the main difference. The primary outcome and assessment point were clearly defined, and the researchers reported effect estimates rather than relying solely on statistical significance.
However, both trials were small: 40 participants in the depression study and 50 in the post-traumatic stress disorder study. The depression trial’s sample size reflected feasibility rather than a formal power calculation. With samples this small, randomisation cannot be expected to reliably balance all important characteristics between groups, and estimates of treatment effects are imprecise. The wide confidence interval around the depression effect also highlights this uncertainty. The advantage was also outcome- and timepoint-specific, so a larger trial must show that it is robust.
Both studies were conducted in specialist research settings using advanced scanning, data analysis and neuronavigation. Their participants, clinical teams and technical resources may differ from those in routine services, limiting how confidently the findings can be applied to everyday clinical practice.
The depression findings apply to a specific high-dose accelerated protocol and cannot automatically be extended to standard once-daily treatment. Further, the study compared imaging-based target selection with one scalp-based method; it does not show superiority over every simpler targeting approach.
The post-traumatic stress disorder trial found support for only one of its two primary biological outcomes. Eligibility required three of the four symptom clusters (including hyperarousal), so some participants had subthreshold symptoms (baseline checklist scores ranged from 10 to 60). Clinical differences were absent immediately after treatment and emerged later, at a follow-up affected by attrition and intervening treatment. These results are promising but less certain than a clear effect at the main post-treatment assessment. Crucially, its design did not isolate the value added by imaging. A standard-target active-treatment group would have been needed for that comparison.
Implications for practice
Do these trials justify routine imaging-guided targeting?
They make a stronger case for testing it, but not yet for making it routine.
The depression trial provides the more direct evidence. Connectivity-guided target selection was associated with greater improvement than a widely used scalp-based method, with 80% responding compared with 60%. For people who have already tried several treatments without sufficient benefit, that difference could be important if confirmed. However, target selection is only one lever; dose, pulse pattern, session spacing and schedule also matter. Even scalp-based targeting achieved a 60% response rate with this intensive protocol, and the authors note it may remain reasonable where high-quality imaging is unavailable.
Nevertheless, changing practice on the basis of 40 participants would be premature. A larger trial should reproduce the effect across several centres and test whether it persists beyond one month. It should also assess functioning, quality of life and cost-effectiveness.
The post-traumatic stress disorder trial offers a complementary message. It suggests that stimulation delivered to an individually selected surface target can influence a deeper threat-related region. But is changing a brain response valuable if patients do not clearly feel better at the same time? Future trials need three groups: active treatment with personalised targeting; active treatment with standard targeting; and sham treatment. This would distinguish whether stimulation has a clinical effect from whether imaging makes it work better.
Implementation is not a minor issue. Personalised targeting requires access to magnetic resonance imaging, a sufficiently long, high-quality functional scan, validated analysis, quality control, specialist staff and neuronavigation. The depression trial used a 41-minute functional scan and the post-traumatic stress disorder trial used under ten minutes, so how much imaging is enough for a reliable target is a key practical question. These requirements could increase costs and waiting times and could reduce access if personalised targeting depends on specialist academic centres, although centralised analysis of locally acquired scans might reduce that barrier.
Future research should therefore examine:
- whether targets are reproducible across scans and centres,
- whether clinical benefits persist,
- which patients are most likely to benefit,
- whether simpler or shorter imaging methods work as well,
- patient acceptability and adverse effects,
- treatment access, and
- whether better outcomes justify the additional cost.
Although we are excited about circuit-based approaches to improve brain stimulation outcomes, we remain cautious about treating sophisticated technology as synonymous with personalised care. If personalised targeting requires additional scanning, analysis, expertise and cost, it needs to earn that complexity by improving outcomes that matter to patients and not simply by producing a more convincing-looking coordinate on a brain scan.
Returning to the faulty-house analogy, the depression study may have identified a better target, while the post-traumatic stress disorder study suggests that a personalised target can influence the intended circuit. Neither yet makes individual brain imaging a requirement for routine care. A better wiring diagram matters only if the lights actually come back on.
Statement of interests
Echo Mei Xu and Julian Mutz were not involved in either study and have no financial relationships with the researchers or technologies discussed.
Editor
Edited by Éimear Foley. ChatGPT assisted with language refinement and formatting during the editorial phase.
Links
Primary papers
Joseph J. Taylor, Marina R. Kare, Dania Haj-Darwish, Emma Jones, Lauren Sanderson, Sanaz Khosravani, Jessica Leach, Leanna Bomer, Natalie Hall, Nicole Chiulli, Elizabeth Steuber, Christopher Lin, William Drew, Stephan T. Palm, Anjali Chandra, Summer B. Frandsen, Anastasia Bekou, Tracy Barbour, Sheena R. Baratono, Irene Gonsalvez, Stanley Lyndon, Frederic L. W. V. J. Schaper, Wei Wang, David Silbersweig, Shan H. Siddiqi, Michael D. Fox (2026). Connectivity- vs Scalp-Based Targeting of Accelerated Transcranial Magnetic Stimulation for Depression: A Randomized Clinical Trial. JAMA Psychiatry, 83(8), 807–817. https://doi.org/10.1001/jamapsychiatry.2026.1100
Sanne J. H. van Rooij, Ryan Langhinrichsen-Rohling, Sean T. Minton, Cecilia A. Hinojosa, Joshua Lukemire, Rebecca Lipschutz, Rebecca Hinrichs, Natalie Merrill, Timothy D. Ely, Kristina Dahlgren, Patlapa Sompolpong, Gregory Job, Patricio Riva-Posse, Paul E. Holtzheimer, Vince D. Calhoun, Joan A. Camprodon, Sheila A. M. Rauch, Nadine J. Kaslow, Kerry J. Ressler, Tanja Jovanovic, William M. McDonald (2026). Personalized fMRI-Guided TMS Targeting the Threat Neurocircuitry in PTSD: A Randomized Clinical Trial. The American Journal of Psychiatry, 183(5), 343–354. https://doi.org/10.1176/appi.ajp.20250749
Other references
Brown, R., Cherian, K., Jones, K., Wickham, R., Gomez, R., & Sahlem, G. (2024). Repetitive transcranial magnetic stimulation for post-traumatic stress disorder in adults. Cochrane Database of Systematic Reviews, 8(8), CD015040. https://doi.org/10.1002/14651858.CD015040.pub2
Cash, R. F. H., Weigand, A., Zalesky, A., Siddiqi, S. H., Downar, J., Fitzgerald, P. B., & Fox, M. D. (2021). Using Brain Imaging to Improve Spatial Targeting of Transcranial Magnetic Stimulation for Depression. Biological Psychiatry, 90(10), 689–700. https://doi.org/10.1016/j.biopsych.2020.05.033
Cash, R. F. H., & Zalesky, A. (2024). Personalized and circuit-based transcranial magnetic stimulation: Evidence, controversies, and opportunities. Biological Psychiatry, 95(6), 510–522. https://doi.org/10.1016/j.biopsych.2023.11.013
Cole, E. J., Phillips, A. L., Bentzley, B. S., Stimpson, K. H., Nejad, R., Barmak, F., Veerapal, C., Khan, N., Cherian, K., Felber, E., Brown, R., Choi, E., King, S., Pankow, H., Bishop, J. H., Azeez, A., Coetzee, J., Rapier, R., Odenwald, N., Carreon, D., … Williams, N. R. (2022). Stanford Neuromodulation Therapy (SNT): A Double-Blind Randomized Controlled Trial. The American journal of psychiatry, 179(2), 132–141. https://doi.org/10.1176/appi.ajp.2021.20101429
National Institute for Health and Care Excellence. (2015). Repetitive transcranial magnetic stimulation for depression. https://www.nice.org.uk/guidance/htg396
Sack, A. T., Cohen Kadosh, R., Schuhmann, T., Moerel, M., Walsh, V., & Goebel, R. (2009). Optimizing Functional Accuracy of TMS in Cognitive Studies: A Comparison of Methods. Journal of Cognitive Neuroscience, 21(2), 207–221. https://doi.org/10. 1162/jocn.2009.21126