Study On Invasive Brain-machine Interfaces For Hand Sensorimotor Function

Somatosensorimotor function plays a crucial role in our daily life.However,amyotrophic lateral sclerosis(ALS),stroke,traffic and other accidents cause many people severely paralyzed and deprive them of somatosensorimotor function.They have to rely on others for daily life.It brings heavy economic,laborious,and psychological burden for patients,their family and society.Brain-machine interface(BMI),which reads out motor intention directly from cerebral cortex and translates it into commands to control external device,is a promising approach to help these patients and has gone through a fast development in the last decade.Restoring the upper limb function is a special focus in its development because of the importance of upper limb function in daily life.A typical BMI for upper limb function restoration allows a tetraplegic patient to control a robotic arm to reach and grasp the bottle on the desk for a drink.The invasive signal-based BMI has made great progress in controlling the end of robotic arm to freely move in three-dimensional space.In terms of performance,it holds huge advantage over non-invasive signal-based BMI.However,to be functionally useful,the invasive BMI is facing a great challenge: the imprecise decoding of hand movement and consequent failure of control over robotic hand.There are two underlying reasons.(1)Hand movement has much more degrees of freedom than three-dimensional arm.Hence,it requires richer effective neural information.But the current BMI could only acquire limited effective neural information.(2)The natural hand movement control relies on sensory feedback to continuously correct the motor output.Thus,the lack of sensory feedback in current BMIs would lead to wrong motor output,which could not be overcome by the advancement of the decoding algorithms.To address the first problem,we found that most related BMI studies only utilized signals from primary somatosensory and motor cortex and thus ignored signals from other cortices also in the hand movement control circuit.Therefore,as the first focus of the thesis,we proposed including signals from posterior parietal cortex into BMIs to augment the effective neural information.As for the second problem,several groups are trying to intracortically stimulate primary somatosensory cortex to generate the natural sensation.Although individual electrode stimulation could roughly yield specific sensation,the completely natural sensation requires the restoration of the functional connectivity between underlying neuronal ensembles in addition to the activation of individual neuronal ensembles.To this end,the first step is to reveal the underlying functional connectivity of primary somatosensory cortex and how it would be changed by intracortical microstimulation.It is the second focus of the thesis.We used invasive cortical signals to study these two questions.First,we recorded and analyzed the stereo-encephalography(SEEG)signal when epilepsy patients perform hand gestures.The high gamma component of SEEG signals from PPC demonstrated selectivity to different gestures.The introduction of SEEG signals from PPC to previously primary somatosensotimotor cortex-based BMI could enhance the gesture decoding performance.In addition,the first activation time across these three cortices presented a temporal sequence: PPC first,primary motor cortex second,and primary sensory cortex last.This result demonstrates,for the first time,that invasive EEG signal from human PPC contains fine hand movement information and could provide complementary information to primary somatorsensorimotor cortex.Second,to investigate the functional connectivity within primary somatosensory cortex(S1)during resting state,we analyzed data recorded by microelectrode arrays implanted in S1 of a tetraplegic patient during rest.We demonstrated a coherent network within S1 at Alpha-Beta band of local field potential(LFP).This coherence could be enhanced by intracortical microstimulation(ICMS).In addition,the further analysis of a sensation task shows a functional network in Theta band.This result sheds a light on how the electrical stimulation should be applied when restoring sensory feedback.