SCD measures how far the coil sits from its target
Scalp-to-cortex distance is the Euclidean distance between the scalp point where the TMS coil is placed and the cortical target beneath it. Because the magnetic field weakens as it travels through scalp, skull and cerebrospinal fluid, a longer distance means less electric field reaches the cortex at the same machine output. The paper cites an approximate rule that each additional millimeter of SCD requires about 3% more TMS output to produce an equivalent field at the motor cortex.
The authors measured SCD for two predefined targets: the left primary motor cortex (M1) and the left dorsolateral prefrontal cortex (DLPFC), the latter localized to the middle frontal gyrus. They reconstructed each participant's scalp and cortical surface from T1-weighted MRI and used neuronavigation to pinpoint coil position, then computed the distance between the scalp site and the corresponding cortical point. They also formed a ratio, SCD DLPFC/M1, which compares how much deeper one target sits relative to the other within the same person.
A new index compares actual output with SCD-adjusted output
Standard rTMS protocols set intensity as a fixed percentage of each person's resting motor threshold (RMT), the lowest intensity that elicits a motor evoked potential of at least 50 μV in at least half of trials. In this trial, active stimulation was delivered at 120% of RMT, 1500 pulses per session, five days a week for three weeks. That fixed-percentage approach ignores the fact that the DLPFC and M1 may sit at different depths, so the same nominal dose can produce different fields at the two targets.
To capture this, the authors defined the SCD-modified output difference (SCD-MOOD): the gap between the output actually delivered in the trial and an output adjusted for the individual's SCD ratio. A larger SCD-MOOD means the delivered dose exceeded what the SCD-adjusted calculation would prescribe. This is a derived dosing index, not a direct measurement of the electric field in the brain, and the paper treats it as a candidate predictor rather than a validated dosing standard.
Responders had deeper left M1 and a lower DLPFC/M1 ratio
Response was defined as at least a 50% reduction in the Cornell Scale for Depression in Dementia (CSDD) total score; remission was a post-treatment CSDD score below 7. Among patients receiving active rTMS, responders had a significantly greater SCD of left M1 than non-responders at week 3 (22.78 vs 16.11 mm, p = 0.001) and at week 12 (21.85 vs 16.28 mm, p = 0.011). They also had a significantly lower SCD DLPFC/M1 ratio at both time points (0.75 vs 0.93 at week 3, p = 0.002; 0.76 vs 0.93 at week 12, p = 0.005).
The left DLPFC distance alone did not separate responders from non-responders, and cortical thickness of either target showed no significant group difference. In the sham group, no SCD metric or ratio differed between responders and non-responders at either time point. That contrast is central: the geometric signal tracked outcomes only when real stimulation was delivered.
Actual output did not separate groups, but adjusted output did
RMTs and the actual TMS outputs delivered were comparable between responders and non-responders, and between remitters and non-remitters, in both treatment arms. So the raw dose a patient received did not by itself explain who improved. Once the authors adjusted output for region-specific SCD, responders in the active group showed lower SCD-adjusted outputs and higher SCD-MOOD scores across all three assessments; no such differences appeared in the sham group.
This is the paper's mechanistic argument: what mattered was not the nominal intensity but how that intensity related to the distance the field had to cross at each target. The authors interpret the pattern as evidence that SCD metrics reflect biologically driven dosing effects rather than generalized placebo phenomena, since sham responses were not captured by the same metrics.
Left M1 distance and SCD-MOOD tracked clinical remission
In exploratory binary logistic regression, clinical remission to active rTMS was significantly associated with the SCD of left M1 (p = 0.038) and with SCD-MOOD (p = 0.042). Partial correlation analysis adjusted for age and sex found that the SCD of left M1 correlated positively with change in RMT (r = 0.551, p = 0.005) and with change in CSDD total score at 3 weeks (r = 0.561, p = 0.004) and 12 weeks (r = 0.501, p = 0.013) in the active group; no significant associations appeared in the sham group.
Using SCD metrics as predictive features, the responder-versus-non-responder classification model reached 92.5% accuracy at week 3 and 86.2% at week 12. The authors report AUC values of 0.84 to 0.925 for the SCD of left M1 and SCD-MOOD, which they describe as outperforming the brain-age model from their earlier work and functional-connectivity predictors reported in other trials. Those comparisons are the authors' framing; this packet does not contain the underlying comparison data.
Multicollinearity and a small sample limit the model
The regression diagnostics flagged strong overlap among predictors: variance inflation factors were 16.64 for SCD-adjusted output and 17.81 for SCD-MOOD. The authors kept these variables on theoretical grounds, but such collinearity makes it hard to attribute predictive weight to any single index. The classification accuracy also comes from a small sample of 55 patients split across active and sham arms, so the reported percentages should be read as internal performance rather than an externally validated benchmark.
Two further constraints sit close to the results. The trial delivered 1500 pulses per session, which the authors note is half the total pulse count of the standard FDA-approved protocol for major depression, and treatment lasted three weeks. The authors frame this as a possible strength, arguing that even a substandard protocol produced better outcomes in patients whose dose exceeded their SCD-adjusted level. The sample also lacked subtype classification, so patients with Alzheimer's-type and vascular neurocognitive disorder were not analyzed separately, even though the two conditions have different atrophy patterns and may produce different SCD profiles.
The practical target is region-specific dose, not raw intensity
The paper's most decision-relevant claim is a comparison, not a single threshold: two patients can receive the same 120% RMT output and still differ in the field reaching left M1 and left DLPFC, because the DLPFC/M1 distance ratio varies between individuals. In this cohort, the average SCD was 17.88 mm at left M1 and 14.85 mm at left DLPFC, and baseline RMT correlated with left M1 SCD (r = 0.359, p = 0.022), consistent with the idea that deeper targets need more output to reach threshold.
The authors therefore question routine use of a uniform 120% RMT protocol in older patients with depression, who may show greater SCD heterogeneity from neurodegeneration. Their proposed direction is to compute region-specific SCD from structural MRI and adjust output accordingly, using SCD and SCD-MOOD as pre-treatment features. The evidence supports that priority as a hypothesis worth testing prospectively; it does not yet establish that SCD-guided dosing improves outcomes, because this study measured prediction and association within one small trial rather than comparing SCD-adjusted dosing against fixed dosing.
Sources used in this answer
Treatment response prediction in major neurocognitive disorder with depression using the scalp-to-cortex distance of TMS targets: implications for personalized dose estimation
Pre-treatment MRI-derived scalp-to-cortex distance metrics of left M1 and the DLPFC/M1 ratio distinguished active rTMS responders from non-responders, with classification accuracy of 92.5% at week 3 and 86.2% at week 12 in 55 older adults with depression and major neurocognitive disorder.