Exploring the Effects of Motor Training with Neurofeedback on Brain Functional Network in Stroke Patients Based on Functional Near-Infrared Spectroscopy

ObjectiveTo investigate the effects of motor training with neurofeedback on brain functional network in stroke patients using functional near-infrared spectroscopy (fNIRS).MethodsTwenty patients with subacute stroke and hemiplegic motor impairment treated at the Affiliated Rehabilitation Hospital of National Research Center for Rehabilitation Technical Aids from November 2023 to March 2024 were initially enrolled, with 5 dropouts. All patients sequentially performed three tasks: resting-state task (RST), limb coupling training task (LCT), and limb coupling with neurofeedback training (LC-NFT). During each task, a 40-channel fNIRS was used to measure oxyhemoglobin (HbO₂) concentration changes in bilateral dorsolateral prefrontal cortex (DLPFC), pre-supplementary motor area (PSMA), primary somatosensory cortex (PSC), primary motor cortex (M1), and occipital lobe (OL). Based on the fNIRS signals collected, phase locking value (PLV) was calculated to assess connectivity between each pair of channels, followed by computing the mean PLV values between regions of interest (ROIs) to evaluate functional connectivity characteristics under different task conditions. Statistical analyses were performed to compare differences in functional connectivity strength across the three task states. Graph theoretical analysis was applied to calculate network metrics, including clustering coefficient, characteristic path length, and global efficiency, for further comparison. Pearson correlation analysis was performed to assess the relationship between clustering coefficient values of bilateral DLPFC and Fugl-Meyer Assessment for Upper (FMA-UE) and Lower Extremities (FMA-LE).ResultsCalculation of PLV value changes across three task states showed that, compared with RST, LC-NFT significantly increased PLV values in cDLPFC-cPSMA, cDLPFC-iPSC, cDLPFC-cM1, cDLPFC-iM1, cDLPFC-iOL, iDLPFC-cPSMA, iDLPFC-iM1, iDLPFC-cOL, iDLPFC-iOL, cPSMA-iPSMA, cPSMA-iM1, cPSMA-cOL, iPSMA-cM1, iPSMA-iM1, iPSMA-cOL, cM1-iM1, iM1-cOL, and cOC-iOL (<italic>P</italic>&lt;0.05). Compared with LCT, LC-NFT showed higher PLV in cDLPFC-cPSMA, cDLPFC-iOL, iDLPFC-cPSMA, iDLPFC-iOL, cPSMA-cOL, cPSC-iOL, and cOL-iOL (<italic>P</italic>&lt;0.05). Clustering coefficients in cDLPFC, iDLPFC, cPSMA, iPSMA, iPSC, cM1, iM1, cOC, and iOC were significantly higher during LC-NFT than in the RST (<italic>P</italic>&lt;0.05), while clustering coefficients of iDLPFC, iM1, and cOC were significantly higher in the LCT than in the RST (<italic>P</italic>&lt;0.05). LC-NFT exhibited longer characteristic path lengths than RST and LCT (<italic>P</italic>&lt;0.05) and lower global efficiency than RST (<italic>P</italic>&lt;0.05). During LC-NFT, clustering coefficients of bilateral DLPFC were negatively correlated with FMA-UE and FMA-LE scores (<italic>P</italic>&lt;0.05), whereas no linear correlations were observed during RST or LCT (<italic>P</italic>&gt;0.05).ConclusionMotor training with neurofeedback effectively promotes reorganization of brain functional networks in stroke patients, demonstrating significant clinical value.