ELECTROCORTICOGRAPHY SIGNALS ON PROSTHETICS IN A HUMAN SUBJECT

robotic hand capturing butterfly

ELECTROCORTICOGRAPHY SIGNALS ON PROSTHETICS IN A HUMAN SUBJECT

Review of Electrocorticography Signals on Prosthetics in a Human Subject

The article “Individual Finger Control of the Modular Prosthetic Limb using High-Density Electrocorticography in a Human Subject” by Hotson, McMullen, Fifer, Johannes, Katyal, Para, Armiger, Anderson, Thakor, Wester, and Crone (2016), employs the use of native sensorimotor depictions of fingers in the brain-machine interface to attain real-time online regulation of distinct prosthetic fingers. According to this article, brain-machine interfaces provide an auspicious technique for reinstating functions to individuals suffering from severe paralysis.  Furthermore, the article argues that by providing direct cortical regulation over robotic prosthetic devices, brain-machine interfaces may allow patients with spinal cord injury (SCI) to carry out activities of daily living (ADLs) vital for their self-sufficiency. Nonetheless, the article agrees that most of these activities, including taking medications and preparing food, necessitate a certain extent of hand dexterity that has not yet been attained by the brain-machine interfaces.

The article states that this ability can only be attained through intricate hand activities based on the regulation of distinct fingers. Deciphering the neural associates of finger regulation, as highlighted by this article, has been analyzed in the past (Fifer, Acharya, Benz, Mollazadeh, Crone, & Thakor, 2012). Nonetheless, Fifer et al. (2012) agree that researchers have not been able to depict online independent regulation of individual fingers. The article acknowledges that online cortical regulation of finger activities has only been attained with regards to the coordinated actions of multiple fingers. According to Hotson et al. (2016), electrocorticography (ECoG) motions obtained from sensorimotor regions have been utilized offline to restructure hand aperture and categorize various hand gestures. Moreover, the article also claims that electrocorticography has also been employed to continually regulate gasping motions in line with arm motions and categorize diverse types of grasps (Yanagisawa, Hirata, Saitoh, Goto, Kishima, Fukuma, Yokoi, Kamitani, & Yoshimine, 2011).

Hotson et al. (2016) also assert that neural firing rates from microelectrode arrays (MEAs) have also been employed to conduct online cortical regulation of grasping in both humans and nonhuman primates. As stated in this article, offline deciphering of finger actions has been attained with microelectrode arrays in nonhuman primates. However, MEA recordings are rare among humans and experience suffering from abrasion of reliable individual units throughout time. Besides, the article asserts that the constrained cortical regions covered by the microelectrode arrays may fail to adequately sample the cortical linkages of potential use in brain-machine interfaces. Hotson et al. (2016) show that the movements of fingers are epitomized in a greater section of sensorimotor in humans compared to monkeys, and despite the potential significant overlap in motor depictions for distinct fingers, it is integrally difficult for microelectrode arrays to leverage what somatotopic is available.

On the other hand, Hotson et al. (2016) assert that electrocorticography can deliver stable increased area coverage of sensorimotor regions, offering control from different anatomical spots. For instance, Hotson et al.’s “Individual Finger Control of the Modular Prosthetic Limb using High-Density Electrocorticography in a Human Subject”  claims that coverage of both hand and arm regions using a single electrocorticography grid allows a subject concurrently to control their capacity to grasp or reach an object with a prosthetic limb. This study, alongside others conducted, have mainly employed the use of electrocorticography high gamma signals for online regulation (Yanagisawa, Hirata, Saitoh, Goto, Kishima, Fukuma, Yokoi, Kamitani, & Yoshimine, 2011). The study agrees that these spectral reactions are strong enough to be sensed in single trials. Nonetheless, because they depict firing rate shifts in populaces of cortical neurons, their functionality relies on the extent of utility segregation among these populaces.

However, despite this evidence, Hotson et al. (2016) also argue that both functional magnetic resonance imaging (fMRI) and electrocorticography studies have discovered some level of separability in the high population reactions for distinct singers in the precentral gyrus. As stated in the article, the individual finger reactions have been utilized to conduct offline reconstruction and the categorization of finger activities from electrocorticography recordings over sensorimotor regions. Furthermore, offline categorization amid four fingers and rest is acclaimed to be close to perfection in one subject according to the studies conducted in the past (Yanagisawa et al., 2011). Nonetheless, a drastic drop is experienced in the performance of the system when it is translated into an online categorization system synchronized to the presentation of cues. Moreover, the article shows that various groups have conducted offline regression of individual finger motions, but none of these offline regressions and categorizations, have been translated into asynchronous online regulation of distinct fingers. This has been a major hindrance among scholars who try to match these online regressions and categorizations into asynchronous online control of different fingers.

As a result, Hotson et al. (2016) have shown that increased-density electrocorticography electrodes over sensorimotor regions can not only distinguish distinct fingers offline but also be utilized to asynchronously categorize and sense finger motions offline. A human subject used in the study showed online regulation of the increasingly dexterous Modular Prophetic Limb (MPL) (Yanagisawa et al., 2011). Moreover, the brain-machine interfaces depended on the cortical reactions of the subject during the movements of his parallel native fingers, without the arbitrary mapping of the subject’s inputs to finger motions or training. Furthermore, the article’s technique of employing the use of the native functional structure of the sensorimotor cortex averts the need for operant conditioning, thus potentially offering real-time instinctive regulation to individuals that can be enlarged to an increased level of degrees of freedom without putting a significant cognitive burden on the person.

To conduct this research, Hotson et al. (2016) used one human subject, who was a twenty-year-old right-handed man. The patient undertook a left craniotomy for the embedding of intracranial electrodes for the purpose of localizing the brain areas accountable for his seizures to control respective surgery. The study comprised an increased-density eight by sixteen assortment of subdural electrodes over sensorimotor areas. Electrocorticography motions were noted using distant standard subdural electrodes with a diameter of 2.3 mm for reference and ground. Before the subject could be used in the study, he provided an informed consent, which required that he be tested based on a protocol ratified by Johns Hopkins Medical Institutions’ Institutional Review Board.

Hotson et al. (2016) began by localizing the electrodes in line with the cortical surface anatomy using volumetric co-registration of the pre-embedding magnetic resonance imaging with their post-surgical computed tomography (CT) through the use of the Bioimage Suite. The idea was meant to guide the selection process of the electrodes for the brain-machine interface. Following the running of the brain-machine interface, the researchers examined the reconstruction against intraoperative images from the elucidation and implantation of the increased density grid. The locations of the electrodes on a dual-dimensional image of the reconstruction were adjusted manually in relation to the underlying cortex through scaling, translation, and rotation of the grid to increase the configuration between the grid and the prominent sulcal and gyral landmarks available in both the intraoperative images and the 3D reconstruction.

To conduct offline experimental testing, a finger tapping activity was conducted to gather training information from the online finger decipherer. Online brain-machine interface testing was conducted the following day. Furthermore, a further finger tapping activity and a passive vibration task were conducted for offline examination. Further, a vibrotactile stimulation experiment was run to examine activation resulting from somatosensory feedback (Hotson et al., 2016). Moreover, from the experiment, it was identified that although the activation of normal people during motor-imagery deafferented subjects fails to show cortical activation straightly prompted by somatosensory feedback. However, the vibrotactile stimulation experiment was used by this study to control the effect.

To conduct online categorizations of finger motions, Hotson et al. (2016) coached a hierarchical categorizer on the offline finger tapping information. In addition, for any particular period, the study used a binary linear discriminant analysis (LDA) to identify if there was some movement. Thus, permitting the research team to decipher the movements of the fingers devoid of matching to any cues. Subsequently, upon detecting any motion in the finger, a categorization from a second LDA categorizer was employed in the identification of which among the subject’s five fingers moved. The binary motion categorizer was branded a first-order Markovian transition probability whereby the prior for likelihood at time t was identified by the likelihood time t-1 (Hotson et al., 2016). The priors were conducted online just prior to the online testing period to discover a perfect tradeoff amid false negatives and false positives.

Following the detection of motion in a finger from the neural information, the modular prosthetic limb finger was instructed to flex at a constant velocity (Hotson et al., 2016). If the finger was not categorized as moving, it was instructed to outspread back towards its resting position at a fixed speed. Prior to the starting of the online testing session, the subject under study was granted a practice session and open use time to acclimate to controlling the modular prosthetic limb fingers. In this acclimation period, the false-positive degree of motion categorizer was tuned by regulating the priors on the Markovian positive likelihoods.

Hotson et al.’s “Individual Finger Control of the Modular Prosthetic Limb using High-Density Electrocorticography in a Human Subject” examined the accuracy of the binary motion detection categorization and the 5-way distinct finger categorizations to analyze the effectiveness of the brain-machine interface. Motion periods relative to trial beginnings were identified from the video examination for trials where the hand of the subject was visible. The degree of motion versus non-motion likelihoods was discarded in 250ms in relation to movement averaged across trials and at the start of the experiment (Hotson et al., 2016).

The offline inspection was conducted to identify the anticipated range of deciphering performance with an enhanced electrode selection during the periods before the somatosensory feedback (Yanagisawa et al., 2011). Although the absence of somatosensory feedback fails to preclude activation of S1, direct activation of the cortex as a result of somatosensory feedback was found as not being present in deafferented individuals requiring brain-machine imaging (Hotson et al., 2016). The study controlled this aspect through the comparison of the time course of individual finger instigation during the finger vibration and finger-tapping tasks.

According to Hotson et al. (2016) and illustrated in supplementary Fig 1 below.  The subject conducted 39 visually cued experiments during the online trials of the MPL fingers. From the results, the patient started moving the impelled finger averagely 1.43 seconds after the signal was prompted. The likelihoods were discovered at the frequency of 24 Hz. The study also used an average of features extracted over 2.34 seconds, with a group delay of about 1.17 seconds. From the offline analysis, a total of 33 electrodes showed significant task-lined spectral variation during the movement of the fingers in the 512 ms based on a motion at the start of the experiment. The electrodes traversed significance verges at a median of 16 ms before the start of the experiment. The twenty-five electrodes with significant variation in the 512 ms based on vibrotactile stimulus onset attained a significance median of 80 ms after the vibration was prompted. The idea showed a significant deal of the initiation of finger movements was not entirely dependent on somatosensory feedback.

ELECTROCORTICOGRAPHY SIGNALS

Hotson et al. (2016) demonstrated, for the first time amongst humans, online neural deciphering of specific finger motions to regulate a dexterous modular prosthetic limb. The study employs the use of increased gamma power obtained from a high-density electrocorticography grid placed over the sensorimotor cortex to create instinctive brain-machine interface finger control. The study’s online outcomes contribute to the increasing literature of online electrocorticography to asynchronously moderate the opening of two grasp categories while regulating elbow motions with the prosthetic limb.  The capability to concurrently regulate grasp and reach autonomously with a prosthetic limb established with mean accuracies of 81 percent and 84 percent for grasp and reach, respectively (Yanagisawa et al., 2011).

Additionally, electrocorticography control signals have been incorporated with intelligent robotics to get hold of and move objects, whereby a success rate of 70 percent is attained. Yanagisawa et al. (2011) managed to asynchronously detect hand movements, whereby 61 percent were detected within a second, and then identified which among the three diverse hand movements was conducted. Fifer et al. (2012) also demonstrated 97 percent accuracy in categorizing between grasping and resting at a fixed period in relation to signals prompted among two individuals.

In conclusion, the analysis presented in this study showed that there is increased pre-movement activity throughout postcentral and precentral gyri during the process of finger tapping. Most of the decoding accuracy attained by this study was obtained from the electrodes that also demonstrated initiation during vibrotactile activation. This may have resulted from the idea that the high-density grid had better coverage of the finger depiction of the postcentral gyrus compared to the precentral gyrus. Another reason could be due to the more discrete somatotopy of separate fingers in the postcentral gyrus in comparison to that in the precentral gyrus. The way the study has been adequately presented, including how it conducts its experiments, illustrates its findings, and employs the use of other previous studies to better its findings indicates that it should be published in its current form.

References

Fifer, M. S., Acharya, S., Benz, H. L., Mollazadeh, M., Crone, N. E., & Thakor, N. V. (2012). Toward Electrocorticographic control of a dexterous upper limb prosthesis: Building brain-machine interfaces. IEEE Pulse3(1), 38-42. doi:10.1109/mpul.2011.2175636 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3987748/

 

Hotson, G., McMullen, D. P., Fifer, M. S., Johannes, M. S., Katyal, K. D., Para, M. P., Amiger, R., Anderson, W. S., Thakor, N. V., Wester, B. A., & Crone, N. E. (2016). Individual finger control of a modular prosthetic limb using high-density electrocorticography in a human subject. Journal of Neural Engineering13(2), 026017. doi:10.1088/1741-2560/13/2/026017 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4875758/

 

Yanagisawa, T., Hirata, M., Saitoh, Y., Goto, T., Kishima, H., Fukuma, R., Yokoi, H., Kamitani, Y., & Yoshimine, T. (2011). Real-time control of a prosthetic hand using human electrocorticography signals. Journal of Neurosurgery114(6), 1715-1722. doi:10.3171/2011.1.jns101421 https://pdfs.semanticscholar.org/d1e4/20996f12b98a598b8292c5fcf033cbbed223.pdf?_ga=2.89786203.1743189148.1595652845-484883021.1586142432

0 Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Subscribe to our newsletter

The Injury Lab © 2020 All Rights Reserved.