Test-retest reliability and symptom association of personalized depression TMS targets: A comparative study of...

Personalized transcranial magnetic stimulation (TMS) targeting holds promise for improving depression treatment, but its clinical translation is hindered by limited open-source implementation and systematic comparisons of target reproducibility and clinical relevance. We implemented two leading personalized TMS-target generating approaches, namely refined seed-based (RSA) and hierarchical clustering (HCA) algorithms, and compared them on (1) test-retest reliability of derived targets, and (2) association of target-sgACC connectivity with depressive symptoms. Using resting-state fMRI data from healthy and depressed individuals, spatial reliability was quantified via inter-run Euclidean distances, and clinical relevance was assessed through correlations between depression severity and functional connectivity of targets with sgACC. Effects of global signal regression (GSR) were also evaluated. The results showed that RSA produced targets in more superior and postrior part of DLPFC and demonstrated significantly higher test-retest reliability than HCA (smaller inter-run Euclidean distances). Further, RSA-derived target-sgACC connectivity correlated positively with depression severity, which was absent in HCA-derived targets. In addition, GSR improved spatial reliability for RSA but not HCA. Our results indicate that RSA exhibits superior test-retest reliability and symptom association compared to HCA, yet large-scale clinical trials are warranted to determine which approach yields superior therapeutic efficacy, and open-sourced implementation may accelerate clinical adoption.