Functional Connectivity Targeting in TMS: Beyond the 5cm Rule

Where on the scalp the treatment coil is positioned during transcranial magnetic stimulation for depression has been a moving target — literally — since the original protocols of the mid-1990s. The earliest method, George and colleagues' "5cm rule" (1995), positioned the coil 5 cm anterior to the motor cortex hand area (M1) along the scalp, an easily reproducible bedside landmark that became the de facto standard for the first generation of clinical trials.¹ A decade of subsequent functional-imaging work has shown that the most effective left dorsolateral prefrontal cortex (DLPFC) targets correlate with anti-correlated resting-state functional connectivity to the subgenual anterior cingulate cortex (sgACC) — a network finding first demonstrated by Fox and colleagues (2012) and refined through individualized, fMRI-guided targeting in subsequent trials.² This article walks through the evolution of TMS targeting from the 5cm rule to scalp-based refinements (BeamF3) to fMRI-guided personalized neuronavigation, and what the literature establishes — and does not — about which method matters for routine clinical care in 2026.

The original 5cm rule and its limitations

The 5cm rule emerged from George and colleagues' early trials of prefrontal rTMS for depression in 1995–1997.¹ The method: locate the motor cortex by finding the lowest stimulator output that produces a visible thumb twitch over the contralateral hand area (the "motor hotspot"), then move the coil 5 cm anteriorly along the parasagittal line. The site so defined was treated as the left DLPFC for purposes of treatment delivery.

The rule's clinical virtue is its reproducibility at the bedside: no imaging required, no specialized equipment beyond the stimulator and a measuring tape. Its weakness is anatomical. The actual cortical region under the coil when the 5cm rule is applied varies substantially between patients because of differences in head size, skull shape, and the precise position of the motor hotspot. Herbsman and colleagues (2009) quantified this variability — patients with depression who responded best to 5cm-derived targeting were those whose stimulation site landed in a more anterior and lateral DLPFC sub-region than the group average, suggesting that some patients were being treated at sites that the rule did not reliably localize.³

The DLPFC is functionally heterogeneous. Adjacent millimeters of cortex serve meaningfully different roles in cognitive control, working memory, and emotional regulation, and the boundary between Brodmann areas 9 and 46 — both of which are sometimes referred to as "the DLPFC" — does not respect scalp landmarks. A targeting method that places the coil within a several-centimeter cloud of possible cortical sites is, in mechanistic terms, treating somewhat different brain regions in different patients.

The sgACC connectivity finding (Fox 2012)

The most consequential refinement to the targeting question came from a network-neuroscience reframing. Fox and colleagues (2012) used resting-state functional MRI data to map the intrinsic functional connectivity of every point in the DLPFC to the subgenual anterior cingulate cortex (sgACC), a deep medial-frontal region implicated in depression.² They then reviewed the published rTMS trials and asked a simple question: did the DLPFC stimulation sites that produced larger antidepressant effects share a common network signature?

The answer was yes. DLPFC sites with stronger anti-correlated functional connectivity to the sgACC produced better antidepressant response across the published literature.² The sgACC is a region whose hyperactivity is one of the most replicated findings in depression neuroimaging — the early Mayberg work on sgACC hypermetabolism is foundational to the network model of depression. The Fox finding provided a network-level rationale for why DLPFC stimulation works at all: a superficial cortical target is engaged because of its anti-correlated connectivity to a deeper limbic node that surface stimulation cannot reach directly. Modulating the DLPFC indirectly modulates the sgACC through the anti-correlated network connection.

That finding reframed the targeting question. The 5cm rule was not wrong so much as imprecise — it sometimes landed on a DLPFC sub-region with strong sgACC anti-correlation, sometimes on a sub-region without. The optimal target is not a fixed scalp coordinate but a patient-specific cortical location defined by network connectivity.

Functional-connectivity-guided targeting

The natural next step was to use individual-patient resting-state fMRI to identify the DLPFC sub-region with the strongest anti-correlation to the sgACC in that specific patient, then to stimulate that personalized site using image-guided neuronavigation. Williams and colleagues (2018) were among the first to test this approach in an accelerated open-label protocol and reported a striking response signal — though in a small, non-randomized sample that the authors themselves framed as hypothesis-generating.⁴

Subsequent work, including the SAINT trial (Cole et al., 2022, Stanford), formalized the approach — fMRI-guided targeting to a personalized DLPFC site, combined with an accelerated treatment schedule of multiple iTBS sessions per day over five consecutive days.⁵ The SAINT findings were notable enough to drive a wave of interest in personalized accelerated TMS; subsequent replication attempts have shown variable results, and the field is still working through which of several methodological features — personalized targeting, accelerated schedule, higher pulse dose, sham comparator design — accounts for the observed effect sizes.

What the literature now supports, hedging appropriately:

• Personalized fMRI-guided targeting is a plausible refinement of DLPFC stimulation, with mechanistic rationale grounded in Fox 2012 and pilot-trial signal in Williams 2018 and Cole 2022.²⁴⁵
• It is not, in 2026, an established requirement for clinically meaningful antidepressant response to TMS. Standard outpatient TMS using less individualized targeting methods continues to produce response rates consistent with the large naturalistic registries.
• Clinical implementation is limited by two practical constraints: the patient must have a resting-state fMRI scan suitable for individualized connectivity mapping, and the clinic must operate image-guided neuronavigation hardware and the analytic pipeline that translates the scan into a stereotactic coordinate. Most outpatient TMS clinics do neither.

The middle ground — BeamF3 and scalp-based refinements

Between the imprecise 5cm rule and the resource-intensive fMRI-guided neuronavigation lies a practical middle ground used by most modern outpatient TMS clinics. The BeamF3 method (Beam et al., 2009) is a scalp-measurement system that approximates the F3 electrode position from the international 10-20 EEG system, adjusting for individual head size and shape using a small number of skull measurements.⁶ The F3 site sits over the left DLPFC and is anatomically more consistent across patients than the 5cm-derived site.

Mir-Moghtadaei and colleagues (2015) compared the cortical localization of the 5cm rule and the BeamF3 method against MRI-defined DLPFC anatomy and found BeamF3 to be more accurate than the 5cm rule at landing the coil over the left DLPFC, with less inter-patient variability.⁷ BeamF3 does not match fMRI-guided personalized targeting on theoretical grounds — it still uses a population-average scalp landmark rather than the patient's own connectivity map — but it is substantially more accurate than the original 5cm rule and requires no imaging.

BeamF3 has been widely adopted in clinical practice. The Clinical TMS Society consensus recommendations (Perera et al., 2016) acknowledge BeamF3 as a reasonable scalp-based targeting method and frame the targeting question as device-agnostic, with the choice among 5cm, BeamF3, and image-guided methods reflecting clinic resources and clinician preference rather than a single mandated standard.⁸

In current outpatient practice, most TMS clinics use one of three approaches: the 5cm rule (still in use, particularly at older clinics), BeamF3 (now common), or MRI-guided neuronavigation (less common, more often at academic centers). True resting-state-fMRI-guided personalized targeting — the approach used in SAINT and the Williams 2018 trial — remains rare outside research settings.

The accelerated-protocol connection

Personalized targeting did not arrive in the literature alone. It arrived bundled with accelerated treatment schedules — protocols that compress the standard six-to-nine-week course into days by delivering multiple iTBS sessions per day. SAINT delivered ten daily sessions over five days using fMRI-guided personalization; Williams 2018 used a similar accelerated framework.⁵⁴

That bundling is methodologically important: the published accelerated-personalized protocols tested both interventions together, and disentangling how much of the observed effect derives from personalized targeting versus from the accelerated schedule (or from the higher total pulse dose delivered) is an active research question. Replication trials of accelerated TMS that have used population-average targeting rather than personalized fMRI guidance have shown response signals but with effect sizes that have varied between sites. The current consensus position — appropriately cautious — is that personalized targeting may be one of several factors contributing to improved outcomes when combined with accelerated schedules, not the sole driver.

The conservative interpretation: accelerated, personalized iTBS is a promising research direction with replication still in progress, not a settled clinical alternative to standard-schedule outpatient TMS.

What this means for patients in 2026

A patient evaluating TMS in 2026 should know:

• Standard outpatient TMS in 2026 is effective. Response rates in routine clinical practice, using BeamF3 or 5cm-derived targeting, approximate 58% with remission near 37% — the benchmark naturalistic figures.⁹ These rates were established with the older targeting methods and remain the baseline for what outpatient TMS delivers.
• fMRI-guided neuronavigation is an enhancement, not a requirement. Patients receiving standard targeting at properly equipped clinics still respond at evidence-based rates. Marketing language that frames personalized targeting as essential overstates the published evidence.
• The research literature continues to refine the targeting question. Personalized connectivity-guided targeting is an area of active research; clinicians who follow the literature should describe it accurately rather than over-claim it. A clinic that uses BeamF3 is not delivering an inferior treatment; a clinic that uses fMRI-guided neuronavigation is delivering an enhancement whose marginal benefit is still being characterized in randomized trials.
• Targeting is one variable among many. Pulse count per session, total course duration, motor threshold calibration, protocol selection (10 Hz rTMS vs iTBS), and adherence to the session schedule all matter. A clinic's targeting method is not, in isolation, the determinant of clinical outcome.

Key takeaways

• The 5cm rule (George 1995) was the original DLPFC targeting method — reproducible at the bedside, but with substantial inter-patient variability in the actual cortical site stimulated.¹³
• Fox and colleagues (2012) demonstrated that DLPFC stimulation sites with stronger anti-correlated resting-state functional connectivity to the subgenual anterior cingulate cortex (sgACC) produced better antidepressant response — the foundational network-level finding behind connectivity-guided targeting.²
• The BeamF3 method (Beam 2009) approximates the F3 EEG site and is more accurate than the 5cm rule at landing the coil over the left DLPFC.⁶⁷ It is now the most common scalp-based targeting method in outpatient practice.
• Individualized, fMRI-guided targeting to the patient-specific DLPFC sub-region with maximal sgACC anti-correlation has shown promise in pilot and small-trial work (Williams 2018; Cole et al., SAINT 2022), most often bundled with accelerated treatment schedules; replication of effect size has been variable.⁴⁵
• The Clinical TMS Society consensus (Perera 2016) treats targeting as device-agnostic, with the choice among 5cm, BeamF3, and image-guided methods reflecting clinic resources and clinician preference.⁸
• Routine outpatient TMS in 2026 generally uses BeamF3 or 5cm-derived targeting and continues to produce response and remission rates consistent with the published naturalistic registries; fMRI-guided neuronavigation is an enhancement, not a requirement for clinically meaningful response.

Patients across Anaheim and Orange County evaluating TMS — particularly those with treatment-resistant depression who have read the research literature and want to understand how their treating clinician thinks about targeting — typically find a brief candidacy consultation clarifies which factors actually drive outcome in their case. Our team handles insurance verification directly before the first session, and the TMS therapy page documents the clinic's approach to protocol selection, candidacy, and the standard six-to-nine-week course.

Sources / Further reading