Effect of Database, Pre-processing Pipeline and Scan Duration on rs-fMRI Functional Connectivity Based Treatment Targets for TMS Obtained Using the Stanford Neuromodulation Therapy (SNT) Protocol

Abstract Background Transcranial Magnetic Stimulation targeting the dorsolateral prefrontal cortex (DLPFC) is a therapeutic approach for treating depressive disorder. However, traditional anatomical methods often fail to localize a treatment target. The functional connectivity method represented by the Stanford Neuromodulation Therapy (SNT) has been proposed and demonstrated efficacy. A step of SNT is to achieve DLPFC target localization using resting-state functional magnetic resonance imaging (rs-fMRI) data, based on measurement of the negative correlation coefficient between the functional connectivity of DLPFC and the subgenual anterior cingulate cortex (sgACC). Methods This study has three goals. Firstly, to apply the SNT protocol to investigate variability in target localization in four groups (Human Connectome Project [HCP], China Human Connectome Project [CHCP], major depressive disorder [MDD] from Huaxi Depression Imaging Cohort Study [HX-DICS], and healthy controls [HC] from HX-DICS) from three databases. Secondly, to investigate potential differences in how the DPABI and FSL rs-fMRI pre-processing pipelines, particularly the spatial smoothing, denoising and temporal filtering steps, affect the localization of individualized DLPFC targets. Thirdly, to assess the effect of rs-fMRI scan duration (ranging from 6 to 30 minutes) on target localization. Results Significant inter-subject variability in DLPFC coordinates was found, emphasizing the necessity of individualized targeting. DPABI and FSL pre-processing produced significant differences in target coordinates, which exceeded 1 cm, with mean Euclidean distances ranging from 1.45 cm to 3.82 cm across cohorts, and individual maximum deviations reaching 6.14 cm. FSL pipeline consistently produced higher negative correlation coefficients than DPABI. Scan duration analysis indicated that after 12 minutes, the stability of individual target positions no longer improved significantly with prolonged scanning time, suggesting that approximately 12 minutes may provide sufficiently reliable data for target computation. Conclusion Taken together, findings demonstrate that advancing precision neuromodulation therapy requires optimized data acquisition, standardized workflows, software-specific validation and optimum scan duration.