Characterization and Decoding the Spatial Patterns of Hand Extension/Flexion using High-Density ECoG

During awake brain surgeries, electrocorticogram (ECoG) was recorded using a high density electrode grid from the motor cortex of two subjects while they were asked to execute spontaneous hand extension and flexion. Firstly, we characterized the spatio-spectral patterns of high-density ECoG during the hand movements. In both subjects, we observed event related desynchronization (ERD) in low frequency band (LFB: 8-32 Hz) and event related synchronization (ERS) in high frequency band (HFB: 60-200 Hz) where HFB-ERS was more spatially localized and movement specific compared to LFB-ERD. In particular, improved spatial resolution of high density ECoG revealed HFB-ERS patterns with distinct timing in different anatomical regions. A few channels located anterior to the central sulcus were associated with HFB-ERS which started several hundred milliseconds prior to the movement onset. Several channels were associated with HFB-ERS which started close to the movement onset. Most importantly, only a small number of channels in the motor cortex regions exhibited long duration ERS which lasted while the subjects maintained their hand posture. A common spatial pattern (CSP) algorithm fused with linear discriminant analysis (LDA) was used to distinguish between hand extension and flexion at different time points based on subband features. ECoG data recorded from the channels located either anterior or posterior to the central sulcus were tested separately in classification. For both subjects, using channels located in motor area, HFB yielded almost 100% classification accuracy within 150-250 ms after the movement onset. The classification accuracies obtained from sensory areas were poor compared to motor areas and lagged the movement onset. These results suggest that spatial patterns of motor cortex captured with high-density ECoG in HFB can effectively drive a neural prosthetic to perform hand flexion and extension.