Frontoparietal-sgACC Complexity Shifted Differently by Hemisphere During Accelerated TMS
A 7T resting-state fMRI preprint in depression found a hemisphere-dependent change in frontoparietal-subgenual anterior cingulate cortex dynamics during accelerated TMS, with the primary interaction reaching statistical significance in the main analysis. The signal was not a simple global increase or decrease in network complexity. Instead, the left and right sgACC appeared to move in opposite directions across time, and the pattern differed between patients receiving stimulation and healthy controls.
Twenty-four patients with current depressive episodes and fifteen healthy controls were scanned at baseline and follow-up. Patients underwent accelerated TMS targeted to the left dorsolateral prefrontal cortex using Beam-F3, while controls did not receive treatment. Investigators paired the same bilateral frontoparietal time series separately with left and right sgACC time series, then quantified joint temporal complexity with bivariate multiscale sample entropy across scales a1 through a10. The resulting complexity index, bMSE-CI, was then tested with a Group x Time x Hemisphere model.
The main three-way interaction for the bMSE-CI was significant, with F(1,37) = 6.043, p = 0.0188, and partial Eta squared = 0.140. In practical terms, the trajectories did not mirror one another across hemispheres. Among patients, the mean left-sided change was -0.286 and the mean right-sided change was +0.124. Among healthy controls, the corresponding changes were +0.323 on the left and +0.112 on the right. Those are small numerical shifts, but they point in a directionally distinct pattern of neural coordination rather than a uniform change in resting-state complexity.
The result fits a biologically plausible model of depression as a disorder of altered large-scale network regulation. The sgACC sits at a critical interface between affective processing and cognitive control, while the frontoparietal network participates in executive coordination and top-down modulation. If neuromodulation alters how tightly these systems track one another over time, entropy-based measures can pick up a change in the structure of that coupling even when conventional metrics miss it. The idea is not that a single number captures the entire effect of treatment. It is that the temporal organization of two networks may become more or less regular as depressive circuitry shifts under stimulation.
The Signal Survived in One Scale, Then Lost Strength in Sensitivity Testing
A deeper look at scale-specific effects made the pattern more selective. Of ten three-way interactions across individual scale factors, only a2 remained significant after Benjamini-Hochberg correction, with q = 0.0435. That narrow survival across multiple scales matters because multiscale entropy is designed to ask whether network dynamics change at different temporal resolutions. A result concentrated at one scale suggests a constrained effect rather than a broad restructuring of temporal complexity.
The analysis also looked for a relationship between bMSE and depressive symptom severity measured with MADRS, both at baseline and in change scores. No association survived the specified false discovery correction. In other words, the imaging measure moved in a treatment-relevant direction, but it did not align cleanly with clinical score change in this small sample. For a biomarker candidate, that gap is consequential. A neural readout can be statistically distinct and still fall short of a symptom-linked metric that would matter in routine care.
A random-intercept-only mixed-model sensitivity analysis softened the claim further. In that model, the three-way interaction did not reach statistical significance, with p = 0.0718. This does not erase the primary finding, but it does show how model specification can alter inference in a modest-sized preprint cohort. The result should be read as exploratory rather than confirmatory, especially because the imaging pipeline, the treatment protocol, and the entropy summary each add analytical complexity.
Why Hemisphere-Specific sgACC Dynamics Matter in Depression
The sgACC has long been tied to depressive circuitry and to response patterns in interventional psychiatry. Left-sided stimulation of the dorsolateral prefrontal cortex is often used to influence networks that communicate with limbic and subcortical affective hubs. A hemisphere difference in fp-sgACC complexity may therefore reflect asymmetric engagement of mood-regulating pathways rather than a nonspecific fMRI fluctuation. In that setting, entropy is attractive because it captures irregularity and coordination across signals, not just mean activation.
Still, the biological interpretation remains provisional. Resting-state dynamics are sensitive to medication status, scanner noise, preprocessing choices, arousal, and the exact timing of follow-up scans. The study’s small cohort also limits the certainty of any pathophysiologic claim. Even a statistically significant interaction in this context should be treated as a candidate observation, not a validated mechanism.
The use of 7T imaging adds another layer of interest. Ultra-high-field MRI can sharpen regional signal and potentially improve measurement of subcortical structures such as the sgACC, where smaller effects might otherwise be lost. That technical advantage may explain why a joint complexity metric could reveal a pattern that ordinary summary measures might miss. It does not, however, guarantee that the same effect will appear across sites, scanners, or patient populations.
What This Preprint Can and Cannot Support Right Now
This preprint offers an early argument that bivariate multiscale sample entropy may track longitudinal neural change during accelerated TMS in depression, especially when analyzed across hemispheres. It does not yet show that the measure predicts who improves, who relapses, or who needs a different treatment path. It also does not establish whether the entropy changes are specific to TMS, to symptom improvement, or to nonspecific time effects.
The strongest reading is narrow: frontoparietal-sgACC coordination changed over time in a hemisphere-dependent way in treated patients, and that pattern differed from untreated controls. The weakest reading would be to treat the signal as a ready-made biomarker. The data do not justify that leap.
Independent replication will need larger samples, more robust reliability estimates, and prospective clinical validation. Future work will also need to determine whether the a2 finding persists outside this dataset, whether the entropy metric is stable enough for repeated use, and whether it can separate treatment response from background variability in depressive illness. Until then, the preprint stands as an analytically interesting but still unproven readout of network dynamics in depression during neuromodulation.