Cortical hemodynamic biomarkers forecast rTMS responsiveness in post-stroke upper limb motor recovery

Abstract BACKGROUND Stroke commonly causes upper limb impairment, significantly reducing quality of life. While low-frequency repetitive transcranial magnetic stimulation (LF-rTMS) over contralesional primary motor cortex (M1) improves arm function, its effectiveness varies, particularly in severe cases. METHODS Stroke patients with upper limb paralysis were recruited for 14-day LF-rTMS. Baseline data, including demographics and stroke details, were recorded. Upper limb motor function was assessed using the upper extremity Fugl-Meyer assessment (UEFM) and the Wolf motor function test. Functional near-infrared spectroscopy (fNIRS) data were collected during fist clenching, and the laterality index (LI) was calculated to assess the asymmetry of hemispheric activation in the brain. A model was developed and validated both internally and externally to predict the responsiveness of LF-rTMS. RESULTS This prospective multicenter study included 111 patients with stroke (training cohort, 62; internal/external validation cohorts, 25/24). After 14-day of LF-rTMS, upper limb motor function improved significantly ( P <0.001), with a response rate of 61.3%. The least absolute shrinkage and selection operator (LASSO) and multivariate logistic regression analyses revealed that the pre-intervention UEFM score ( P =0.015, odds ratio [OR]=1.048) and pre-intervention LI for M1 for the affected hand (AFF-LI-M1-pre) ( P =0.001, OR=0.117) were independent predictive factors for the responsiveness of LF-rTMS. The predictive nomogram integrating pre-intervention UEFM score and AFF-LI-M1-pre demonstrated robust discriminative accuracy across cohorts, with AUC values of 0.861 (95% CI 0.770–0.952) in training, 0.853 (95% CI 0.703-1.000) in internal validation, and 0.828 (95% CI 0.645-1.000) in external validation cohorts. Model calibration showed excellent agreement between predicted and observed outcomes (Hosmer-Lemeshow P =0.801). Decision curve analysis (DCA) confirmed clinical utility within broad probability thresholds (0.02-0.94), yielding net benefit advantages of 7%-38% over default strategies. CONCLUSIONS LF-rTMS improved upper limb motor recovery in patients with stroke, with the baseline UEFM score and fNIRS-derived AFF-LI-M1-pre serving as robust predictors. Validated nomogram personalizes therapy, highlighting fNIRS’s role in optimizing treatment. These findings support precision rehabilitation approaches for improving outcomes in resource-limited settings. GRAPHIC ABSTRACT A graphic abstract is available for this article.