Decoding grasp movement from monkey premotor cortex for real-time prosthetic hand control

HAO YaoYao 0 2 3 ZHANG QiaoSheng 0 2 3 ZHANG ShaoMin 0 2 3 ZHAO Ting 2 WANG YiWen 2 CHEN WeiDong 1 2 ZHENG XiaoXiang 0 2 3 0 Department of Biomedical Engineering, Zhejiang University , Hangzhou 310027, China 1 College of Computer Science, Zhejiang University , Hangzhou 310027, China 2 Qiushi Academy for Advanced Studies, Zhejiang University , Hangzhou 310027, China 3 Key Laboratory of Biomedical Engineering of Ministry of Education, Zhejiang University , Hangzhou 310027, China Brain machine interfaces (BMIs) have demonstrated lots of successful arm-related reach decoding in past decades, which provide a new hope for restoring the lost motor functions for the disabled. On the other hand, the more sophisticated hand grasp movement, which is more fundamental and crucial for daily life, was less referred. Current state of arts has specified some grasp related brain areas and offline decoding results; however, online decoding grasp movement and real-time neuroprosthetic control have not been systematically investigated. In this study, we obtained neural data from the dorsal premotor cortex (PMd) when monkey reaching and grasping one of four differently shaped objects following visual cues. The four grasp gesture types with an additional resting state were classified asynchronously using a fuzzy k-nearest neighbor model, and an artificial hand was controlled online using a shared control strategy. The results showed that most of the neurons in PMd are tuned by reach and grasp movement, using which we get a high average offline decoding accuracy of 97.1%. In the online demonstration, the instantaneous status of monkey grasping could be extracted successfully to control the artificial hand, with an event-wise accuracy of 85.1%. Overall, our results inspect the neural firing along the time course of grasp and for the first time enables asynchronous neural control of a prosthetic hand, which underline a feasible hand neural prosthesis in BMIs. - The loss of the hand results in a serious reduction of the functional autonomy of a person in his daily living. Prosthesis is the most common way to restore the lost function, however, there are several barriers existing trough a successful prosthesis use, and the most important seems regarding the development of a reliable interface capable to decode the intention of the disabled to the prosthetic device. Brain machine interfaces (BMIs) provide a new hope for restoring motor functions of the severely disabled through controlling prostheses with intentional commands extracted from brain signals. Decoding motor cortex activities for robotic arm and screen cursor control in two or three dimensions has been examined successfully in human or non-human primates in past decades [13]. However, there are few investigations of movement decoding for restoring hand function, which is a great challenge with a higher degrees of freedom (DoFs). The hand is a marvelous example of how a complex biomechanism can be implemented, with effective combinations of mechanisms, sensing, actuation and cortical control system coordinated in 38 muscles and 22 DoFs [4]. With these complex architectures, recent studies have shown that how dexterous grasps are represented and transformed into motor commands in distinct brain regions. Several cortical The Author(s) 2013. This article is published with open access at Springerlink.com areas distributed in parietal and frontal lobe are involved in control of reaching and grasping movement, such as anterior intraparietal (AIP) area [5], areas PF and PFG of the inferior parietal lobule (IPL) [6] and ventral premotor cortex (PMv, area F5) [7], which form the a dorsal stream pathway in inferior parietal area; some other areas such as area V6A, dorsal premotor cortex (PMd, area F2) [8] and medial intraparietal (MIP) forms another ventral stream pathway [9]. Moreover, these cortical areas are anatomically and functionally interconnected, forming a grasping neural network to complete the function of sensory-motor transformations (both visual and somatosensory information), appropriate hand configuration, grasp movement execution, and highorder motor perceptions [10]. The complexity of such a network, mingled with sophisticated bio-mechanism of reaching and grasping, has kept us from decoding every DoF for grasp under current techniques. Recently, there are some reports that can decode a large number of joint kinematics from hand. Aggarwal et al. tried to decode the individual and two combined fingers flexion and extension movement from neurons in M1, and got high accuracies [11]. Vargas-Irwin et al. demonstrated that the full arm joint kinematics (including arm, wrist and hand) can be reconstructed from local ensembles of M1 [12]. However, none of them can reconstruct the functional grasp gesture, which is crucial for grasping different objects [11,12]. Alternatively, current studies often resolve the problem by transferring continuous grasp movement into discrete classification. The strategy classifies finger configurations into one of the predefined categories based on the kinematic synergy movement in grasping and the neural encoding of grasp postures [13,14]. Grasp types were decoded successfully from multiunit activity (MUA) in PMd and PMv [15], single neuron recording in PMv [16], and multiple units in PMv and AIP [17] previously. Compared with continuous kinematic decoding, the classification strategy takes advantage of simplified hand configuration and high order planning in the brain, reducing the burden of the decoding system dramatically. Current state of arts has specified the grasp related brain areas and some offline decoding results. Townsend et al. demonstrated the first real time grasp types decoding, although it is in synchronous mode and the grasp gestures are induced by different LED light mode, not different object shapes, which are more nature [17]. Hendrix et al. investigated the signaling of grasp dimension and force during reach and grasp movement using signals from dorsal premotor cortex (PMd). However, online decoding grasp movement and real-time neuroprosthetic control using PMd signals have not been systematically investigated. This study, using the discrete classification strategy, presents our work on asynchronously decoding of four gestures and a resting state using neural ensemble signals from the dorsal premotor cortex of a monkey. To obtain the data, we have developed an experimental paradigm for the monkey to grasp one of the four given objects with a specific gesture. Neural signals from the premotor cortex were recorded synchronously with hand movement during the grasp experiment. Individual neuron analysis and population decoding of the signals demonstrated that it is feasible to predict a variety of gestures in real time from the activities of motor cortex. Furthermore, the real-time decoding results were used to control an artificial hand for the first time to achieve the same (...truncated)