Stroke frequently impairs both proprioception and motor control, yet the cortical sensory mechanisms that underlie these deficits remain incompletely characterized. Prior electrophysiological studies commonly quantified the average magnitude of unilateral cortical sensory responses to peripheral vibration, which may miss distributed, trial-to-trial, and bilateral features of sensory processing that are relevant to functional proprioception and motor performance. This study aimed to characterize bilateral cortical sensory responses and determine their relationships with finger proprioception and multiple measures of motor function in people with chronic stroke.
The investigation enrolled forty-six individuals with chronic stroke. During EEG recording, participants received rapid, passive vibrotactile stimulation applied to the left and right fingertips. The paradigm focused on brief, repeated tactile events to evoke somatosensory responses while minimizing active movement to isolate sensory cortical processing.
Cortical sensory processing was quantified using two conventional EEG-derived metrics and one decoding-based measure. Somatosensory evoked potentials (SEPs) were measured at the contralateral parietal cortex to evaluate response magnitude. Event-related desynchronization (ERD) in relevant frequency bands was computed as an additional index of sensory cortical activation. In addition to these unilateral measures, the authors implemented a somatosensory decoder that used combined bihemispheric response patterns to classify which hand (paretic vs. non-paretic) had been stimulated. Decoder performance was defined as the accuracy in identifying stimulated hand from the distributed cortical responses.
Finger proprioceptive performance was assessed with a passive, robotic finger crossing identification task designed to quantify position-sense error without requiring active movement. Motor function was evaluated with three common clinical measures: the Box and Block Test (gross manual dexterity), the Fugl–Meyer Assessment (upper extremity motor impairment), and the Nine Hole Peg Test (fine manual dexterity). These behavioral measures were used to test associations between EEG sensory metrics and both proprioceptive and motor abilities.
When the paretic hand was stimulated, EEG showed reduced ipsilesional SEP and ERD magnitudes compared with stimulation of the non-paretic hand. Thus, unilateral measures reflected attenuated cortical responses on the lesioned hemisphere in response to tactile input from the affected hand.
Two main associations between cortical sensory measures and behavior were reported. First, decreased SEP magnitude was associated with greater finger proprioceptive error on the robotic identification task. Second, reduced somatosensory decoder performance — that is, lower discriminability of which hand was stimulated from bihemispheric patterns — was also associated with increased proprioceptive error.
Importantly, while unilateral response magnitude had relationships with proprioception, the decoder's discriminability of bihemispheric somatosensory patterns showed a stronger relationship with motor performance. The somatosensory decoder explained approximately 22.5% of the variance in motor function across the assessments used, indicating that distributed and lateralized cortical sensory patterns provide information about motor capability beyond contralateral magnitude alone.
The authors applied an exploratory analysis combining multiple motor measures using nonlinear dimensionality reduction and clustering. This approach identified distinct subgroups of participants based on motor assessment profiles. These subgroups showed significant differences in somatosensory decoding performance, suggesting that patterns of distributed cortical sensory responses align with clinically meaningful motor phenotypes after stroke.
The hemispheric distribution and discriminability of cortical sensory responses after stroke are functionally relevant markers of sensorimotor integrity. Assessing relative lateralization and bihemispheric discriminability of somatosensory responses for each hand, rather than focusing solely on the magnitude of the dominant contralateral response, may better capture the reliability of sensory processing and the distributed cortical reorganization that supports sensorimotor function.
These findings support the potential value of a decoding-based neurophysiological measure as a biomarker for sensory-driven rehabilitation, patient stratification, and possibly for guiding targeted interventions. The authors propose that decoder-based assessments of bihemispheric somatosensory patterns could complement traditional EEG metrics when evaluating sensory contributions to motor impairment.
This report is a preprint and has not been peer reviewed. Details about specific decoder algorithms, cross-validation procedures, and exact statistical effect sizes beyond the reported ~22.5% explained variance were not presented in this summary; readers should consult the full preprint for methodological specifics. The authors declared a competing interest: one author has a financial interest in companies that develop rehabilitation devices, with institutional review of those arrangements. Funding sources included the National Institute of Biomedical Imaging and Bioengineering, the National Institutes of Health, the Spinal Cord Injury Research Board (NYSDOH), Albany Stratton VA Medical Center, and others as stated in the preprint metadata.
(Notes: All findings and statements above reflect the content reported in the preprint. The manuscript is posted July 27, 2026, and further methodological or numerical detail should be verified in the full article.)