fNIRS during robot-assisted stroke rehab: training modes and cortical strategies (2026)

• A Frontiers in Human Neuroscience study used 48-channel fNIRS during robot-assisted upper-limb training after stroke across four modes in a single session: Passive, Assistive-active, Active, and Mirror.¹
• Forty-one stroke patients were grouped by functional level, chronicity, and hemiplegic side.¹
• There were no statistically significant differences in global mean activation intensity across subgroups (p > 0.05).¹
• The paper frames the main result as differences in neural strategy (rather than simply higher/lower mean activation) across conditions/groups, with implications for choosing or adapting therapy mode by stroke profile.¹
• The robot used was the Wisebot-X5, a 3D upper-limb exoskeleton (Shenzhen Huaquejing Medical Technology); modes ranged from fully passive trajectory replay at 15°/s to active resistance (1 kg constant load) and mirror therapy via body-sensing camera tracking the unaffected limb at up to 60°/s.¹
• The key finding is a “mode-sensitive” vs. “mode-consolidated” cortical signature: subacute, lower-functioning, and left-hemiplegic patients showed widespread, mode-dependent activation differences across nearly all ROIs (M1, PMC, SMA, DLPFC); chronic, higher-functioning, and right-hemiplegic patients showed a more uniform pattern with fewer inter-mode differences.¹
• Left-hemiplegic patients showed the most extensive cortical recruitment under assistive-active and active modes (all ROIs significantly activated, p < 0.05) but essentially no activation under passive mode; right-hemiplegic patients showed more consolidated and localized activation, particularly under mirror mode (limited to left M1, left FEF, right SMA, bilateral PSC, and SAC).¹
• Clinical translation: the authors propose “mode-sensitive” patients benefit from diversified training starting with Mirror Mode to establish motor intent, then progressing to Assistive-Active Mode; “mode-consolidated” patients benefit from high-intensity, repetitive Active Mode with progressive resistance increases.¹
• The 48-channel NirSmart system (730 nm and 850 nm, 10 Hz sampling) covered bilateral motor-related cortices including M1, SMA, PMC, PSC, DLPFC, FEF, and SAC; 3D digitizer coordinates were registered to MNI standard space.¹
• “Mode sensitivity” is proposed as a more clinically actionable rehabilitation biomarker than simple hemispheric balance indices, shifting the focus from how much the brain activates to how it reorganizes across task demands.¹

Footnotes

1. https://www.frontiersin.org/articles/10.3389/fnhum.2026.1767418 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰