Brain-Machine Interfaces Human-Machine Interfaces (HMI) Reveals That the Amount of Useful Information That Can Be Ex- Josep Miquel Jornet1, Michal K
Total Page：16
File Type：pdf, Size：1020Kb
B mation. A conceptual analysis of the existing Brain-Machine Interfaces human-machine interfaces (HMI) reveals that the amount of useful information that can be ex- Josep Miquel Jornet1, Michal K. Stachowiak2, changed between humans and machines is not and Sasitharan Balasubramaniam3 limited by the capabilities of the human brain 1Ultra-broadband Nano Communication and or those of the machine processor, but by the Networking Laboratory, University at Buffalo, interfaces between them. Simply stated, from an The State University of New York, Buffalo, NY, engineering perspective, the human being can be USA modeled as a macro-system with a processing 2Department of Pathology and Anatomical powerhouse, i.e., the brain, and a collection of Sciences, University at Buffalo, The State peripherals, i.e., the sense organs. All the periph- University of New York, Buffalo, NY, USA erals have their own latency limitations, which 3Telecommunication Software and Systems mainly arise from the fact that they convert a Group, Waterford Institute of Technology, collection of nano-/micro-events, i.e., neuronal Waterford, Ireland activity in the form of action potential signals, into a macro-sized effect, e.g., moving the fingers to touch a display or type some characters in Definition order to control an external machine or, recip- rocally, convert a macro-size effects, e.g., an Brain-machine interfaces (BMIs) refer to com- auditive or visual command, into a a collection munication systems between the brain and an ex- of nano-/micro-events in the brain. ternal device. Desired properties of BMIs include In order to overcome the limitations of the bidirectionality, high spatial and temporal resolu- peripheral system, direct communication with the tion, low invasiveness, accuracy, and robustness. brain is needed. This form of interaction is known In this paper, the different types of BMIs, the state as a brain-machine interface (BMI). Over the of the art, and the future directions are discussed, years, numerous applications have resulted from in addition to highlighting their key applications. these two-way interactions, where compensations are made between the computational capabilities of both systems. Such compensations can be in Historical Background the form of either the computing system analyz- ing the brain signals and adapting the computing For many decades, the interaction between hu- environment or the computing system provid- mans and machines has been restricted to the ing added computing power to compensate for exchange of visual, auditory, and tactile infor- shortcomings of the brain. In the first case, brain © Springer Nature Switzerland AG 2018 X.(S.) Shen et al. (eds.), Encyclopedia of Wireless Networks, https://doi.org/10.1007/978-3-319-32903-1_226-1 2 Brain-Machine Interfaces signals are used to understand the user’s context rent approaches to optogenetic neural interfaces which is then used to control machines, such as include the use of optical fibers coupled to lasers controlling the vehicle. In the second case, this or light-emitting diodes (LEDs) (Zorzos et al. could come in the form of computing systems 2010) and micro-LED arrays (McGovern et al. controlling neuroprosthetic devices for disabled 2010). Moreover, optogenetics enables bidirec- patients. tional interfaces, as light can be utilized both to control and to measure neuronal activity (Kwon et al. 2014). However, the size of existing optical Electrical Brain-Machine Interfaces devices makes them invasive, difficult to contact to individual neurons, and, ultimately, not suit- The most common to date BMIs rely on the col- able for chronic BMIs (Marblestone et al. 2013). lection and excitation of electrical signals from the brain. Electroencephalogram (EEG) signals have been successfully utilized to directly con- trol machines without the need of the sense or- Wireless Brain-Machine Interfaces gans (Millan et al. 2004). EEG signals can be collected in a noninvasive way, i.e., from outside In order to overcome the limitations of tradi- the brain, and support high-temporal resolution, tional electrical and optical BMIs, wireless BMIs i.e., down to the sub-millisecond scale, but have are being developed. In Seo et al. (2013), the limited spatial resolution, i.e., cannot be utilized concept of neural dust was introduced for the to read the action potential signal from a single first time. In the envision architecture, miniature neuron at a time, and are vulnerable to electrical electronic devices or dust motes are implanted artifact sources. Besides EEG-based BMIs, there in the cortex. These devices, which integrate are other more invasive mechanisms that could piezoelectric energy harvesting systems powered be utilized to enable more robust electrical BMIs, by ultrasounds, record the neural activity from the such as intracranial EEG (Leuthardt et al. 2006), cortex and transmit the information to a subdural which is also known as electrocorticogram, and transceiver mounted under the skull. This device microelectrode arrays, which are placed directly is in charge of controlling the neural dust and on the exposed surface of the brain (Hochberg to communicate with the external head-mounted et al. 2006). However, besides their invasiveness, transceiver, where the data is collected. Despite they suffer from several limitations, such as com- the advantages of this wireless architecture, the plex application or unsuitability for long-term fact that it relies on the principles of electrical use. BMIs limits its applications. Recently, in Wirdatmadja et al. (2017), the first wireless BMI based on wireless optogenetic Optical Brain-Machine Interfaces nanonetworking devices (WiOptNDs) was proposed (Fig. 1). WiOptND enables accurate, In parallel to the development of the aforemen- robust, high-throughput, and minimally invasive tioned approaches, the field of optogenetics, i.e., BMIs by leveraging the state of the art in the use of light to interact with genetically modi- nanophotonics, nanoelectronics, and wireless fied neurons in the brain (Zemelman et al. 2002; communication. The fundamental idea is to Deisseroth 2011; Zhang et al. 2007), has experi- replace existing micro-LED arrays and micro- enced a major revolution in the last decade. Op- photodetector arrays used in optical BMIs by tical neural stimulation is considered to be more a network of coordinated nano-devices, which beneficial than electrical neural stimulation, be- are able both to excite individual neurons and cause it permits activation or inhibition of specific to measure their activity. In this application, a types of neurons with sub-millisecond temporal network of collaborative WiOptNDs is utilized precision and eliminates electrical artifacts. Cur- both to excite multiple neurons according to Brain-Machine Interfaces 3 B Skull Control data Ultrasound signals Neuronal stimulating Intermediate model and Transceiver pattern External Transceiver WiOptND Brain-Machine Interfaces, Fig. 1 The WiOptND archi- to generate different neuron excitation patterns as well tecture consists of (i) a network of coordinated nano- as acoustically powering them; and (iii) an external devices able to optogenetically excite and measure the transceiver in charge of acoustically powering the inter- response of neurons; (ii) an intermediate transceiver in mediate transceiver and interfacing it with the actual BMI charge of both controlling the nano-devices in order user incoming commands and to collect, process, and invasiveness of this approach when compared to transmit accurate neuronal activity in real time. existing optogenetic approaches, which require Each nano-device is equipped with an optical bulky lasers or optical fibers. Moreover, by op- nano-transceiver (Feng et al. 2014) and nano- erating at optical frequencies, much higher tem- antenna (Nafari and Jornet 2017), which is able poral resolution than traditional electrical BMIs to both emit and detect optical radiation at a can be achieved. For example, while the main pre-established frequency or wavelength. As features of action potential signals are in the mil- in Seo et al. (2013), WiOptNDs are acoustically lisecond scale, the possibility to measure those powered and remotely controlled through the signals with much higher temporal resolution, subdural transceiver (Fig. 1). such as a few microseconds or even less, may Many benefits in this approach exist. First, the unveil new high-frequency time transients in the very small size of optical nano-antennas (Dorf- action potential signal propagation, which could muller et al. 2010; Nafari and Jornet 2017), below shine new light into the exploration of neuronal 1 micrometer in the largest dimension, enables pathways. This also enables potentially much the possibility to measure the neuronal activ- faster BMIs. For the time being, however, electri- ity in a single neuron, with very high accu- cal and optogenetic BMIs are at an early stage, in racy. Moreover, the total size of each individual which some of the system components have been nano-device, up to a few cubic micrometers at developed and tested, but a fully functional BMI most (Akyildiz and Jornet 2010), minimizes the has not been realized. 4 Brain-Machine Interfaces Future Directions lection from over the skull) have been pri- marily tested in vitro cell cultures or animals. To enable practical long-term implantable wire- One of the promising approaches relies on less BMIs, there are several bottlenecks that need the use of cerebral organoids, i.e., artificially to be overcome.