[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Sign in to use this feature.

Years

Between: -

Subjects

remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline
remove_circle_outline

Journals

Article Types

Countries / Regions

Search Results (55)

Search Parameters:
Keywords = neuroprosthetic

Order results
Result details
Results per page
Select all
Export citation of selected articles as:
16 pages, 22416 KiB  
Article
A Combinatory Therapy of Metformin and Dexamethasone Reduces the Foreign Body Reaction to Intraneural Electrodes
by Bruno Rodríguez-Meana, Jaume del Valle and Xavier Navarro
Cells 2024, 13(24), 2112; https://doi.org/10.3390/cells13242112 - 20 Dec 2024
Viewed by 309
Abstract
Neural electrodes used for bidirectional communication between the nervous system and external devices like prosthetic limbs have advanced in neuroprosthetic applications. However, their effectiveness is hindered by the foreign body reaction, a natural immune response causing inflammation and fibrosis around the implanted device. [...] Read more.
Neural electrodes used for bidirectional communication between the nervous system and external devices like prosthetic limbs have advanced in neuroprosthetic applications. However, their effectiveness is hindered by the foreign body reaction, a natural immune response causing inflammation and fibrosis around the implanted device. This process involves protein adsorption, immune cell recruitment, cytokine release, and fibroblast activation, leading to a fibrous capsule formation and a decrease in electrode functionality. Anti-inflammatory and antifibrotic strategies have the potential to diminish the impact of the foreign body response. In this work, we have evaluated long-term metformin administration and short-term dexamethasone administration as a combined therapy to modulate the foreign body reaction induced by a polyimide intraneural implant in the sciatic nerve of rats. After a 12-week implant, the foreign body reaction was significantly reduced only in the group administered both drugs. Full article
(This article belongs to the Section Cells of the Nervous System)
Show Figures

Figure 1

Figure 1
<p>Results of the functional tests in rats with a PI device implanted in the tibial nerve. (<b>A</b>) Algesimetry test results expressed as percentages of force thresholds for withdrawal (vs. contralateral control paw) of animals before the implantation and after the implantation and treatments for 12 weeks. (<b>B</b>) The plot of the SFI obtained in the walking track test. No significant differences were found.</p>
Full article ">Figure 2
<p>Results of the functional tests in rats with a PI intraneural device implanted in the tibial nerve. Motor nerve conduction parameters of animals before implantation (Pre) and after the implantation of PI devices for 12 weeks and drug administration. (<b>A</b>,<b>B</b>) CMAP amplitudes of GM (<b>A</b>) and PL (<b>B</b>) muscles. (<b>C</b>,<b>D</b>) CMAP onset latencies of GM (<b>C</b>) and PL (<b>D</b>) muscles. No significant differences were found in electrophysiological test results.</p>
Full article ">Figure 3
<p>The effect of drug administration on the FBR to intraneural implants. (<b>A</b>) The number of inflammatory Iba1+ cells in the tibial nerve of animals implanted with PI devices and administered metformin, dexamethasone, or both. (<b>B</b>,<b>C</b>) Tissue capsule thickness around the devices in the tibial nerve of animals implanted with PI receiving the different treatments. Measurements were made using immunofluorescence sections (<b>B</b>) and thin sections of epon-embedded nerves (<b>C</b>). (<b>D</b>–<b>F</b>) Correlation between the number of Iba1+ cells and capsule thickness (IF) at 2, 8, and 12 weeks after implantation. The solid lines represent the linear regression, while the shaded area represents the 95% CIL. * <span class="html-italic">p</span> &lt; 0.05, ** <span class="html-italic">p</span> &lt; 0.01, *** <span class="html-italic">p</span> &lt; 0.001, **** <span class="html-italic">p</span> &lt; 0.0001, and ### <span class="html-italic">p</span> &lt; 0.01 time variable, two-way ANOVA followed by Tukey’s multiple comparison test.</p>
Full article ">Figure 4
<p>Representative images of inflammatory cells (red, Iba 1+ cell) infiltrating the tibial nerve after 2, 8, and 12 weeks of the PI intraneural device implantation in the different groups studied. Note the intense fluorescence emitted by the PI. The area limited by the dotted line corresponds to the tibial fascicle of the sciatic nerve that was used to analyze the number of labeled cells. Scale bar: 100 μm.</p>
Full article ">Figure 5
<p>Representative images of nerve cross-sections around the PI intraneural implant after 2, 8, and 12 weeks of the implantation in the different groups studied. Nerve fibers are labeled with antibody RT97. Note the intense fluorescence emitted by the PI. The measured capsule surrounding the PI device is the area delimited by the dotted line, which separates the PI from the nerve fibers, excluding tissue-empty regions. Scale bar: 50 μm.</p>
Full article ">Figure 6
<p>Representative images of cross-sections of the nerves embedded in epon resin and stained with toluidine blue, corresponding to samples taken at 2, 8, and 12 weeks for the different study groups. The images show the PI implants (pointed to by a red arrow in the top-right panel) within the nerve, surrounded by the capsule and axons. The thickness of the capsule from the implant to the first axons is marked with a red bar in the top-right panel. Images were acquired and transformed to greyscale. Scale bar: 50 μm for all the panels.</p>
Full article ">Figure 7
<p>Representative images of the capsule composition around the PI intraneural implant. Immunohistochemical labeling for macrophages (red, Iba 1+), fibroblasts (green, CD90, arrowheads), and nuclei (blue, DAPI) of tibial nerves of animals of the different groups implanted with a PI device after 2, 8, and 12 weeks. Scale bar: 10 μm. Images with the individual channels are presented as <a href="#app1-cells-13-02112" class="html-app">Supplementary Materials Figures S1–S3</a>.</p>
Full article ">Figure 8
<p>Representative images of nerve sections stained with Masson’s trichrome stain, showing the deposition of collagen in the capsule around the PI intraneural implant. At 2 weeks, the pink-stained area, outlined by the dotted line, corresponds to macrophages around the implant. At 8 and 12 weeks, the pink areas around the devices decreased, while the blue-stained areas (dotted line), composed of collagen fibers, were more preeminent surrounding the implant. Scale bar: 50 and 20 μm.</p>
Full article ">
27 pages, 2015 KiB  
Article
Developing Innovative Feature Extraction Techniques from the Emotion Recognition Field on Motor Imagery Using Brain–Computer Interface EEG Signals
by Amr F. Mohamed and Vacius Jusas
Appl. Sci. 2024, 14(23), 11323; https://doi.org/10.3390/app142311323 - 4 Dec 2024
Viewed by 639
Abstract
Research on brain–computer interfaces (BCIs) advances the way scientists understand how the human brain functions. The BCI system, which is based on the use of electroencephalography (EEG) signals to detect motor imagery (MI) tasks, enables opportunities for various applications in stroke rehabilitation, neuroprosthetic [...] Read more.
Research on brain–computer interfaces (BCIs) advances the way scientists understand how the human brain functions. The BCI system, which is based on the use of electroencephalography (EEG) signals to detect motor imagery (MI) tasks, enables opportunities for various applications in stroke rehabilitation, neuroprosthetic devices, and communication tools. BCIs can also be used in emotion recognition (ER) research to depict the sophistication of human emotions by improving mental health monitoring, human–computer interactions, and neuromarketing. To address the low accuracy of MI-BCI, which is a key issue faced by researchers, this study employs a new approach that has been proven to have the potential to enhance motor imagery classification accuracy. The basic idea behind the approach is to apply feature extraction methods from the field of emotion recognition to the field of motor imagery. Six feature sets and four classifiers were explored using four MI classes (left and right hands, both feet, and tongue) from the BCI Competition IV 2a dataset. Statistical, wavelet analysis, Hjorth parameters, higher-order spectra, fractal dimensions (Katz, Higuchi, and Petrosian), and a five-dimensional combination of all five feature sets were implemented. GSVM, CART, LinearSVM, and SVM with polynomial kernel classifiers were considered. Our findings show that 3D fractal dimensions predominantly outperform all other feature sets, specifically during LinearSVM classification, accomplishing nearly 79.1% mean accuracy, superior to the state-of-the-art results obtained from the referenced MI paper, where CSP reached 73.7% and Riemannian methods reached 75.5%. It even performs as well as the latest TWSB method, which also reached approximately 79.1%. These outcomes emphasize that the new hybrid approach in the motor imagery/emotion recognition field improves classification accuracy when applied to motor imagery EEG signals, thus enhancing MI-BCI performance. Full article
(This article belongs to the Section Applied Neuroscience and Neural Engineering)
Show Figures

Figure 1

Figure 1
<p>The timing diagram for the motor imagery tasks.</p>
Full article ">Figure 2
<p>The embraced ER/MI hybrid pipeline.</p>
Full article ">Figure 3
<p>Clustered column chart visualizing the results generated from the hybrid ER/MI pipeline.</p>
Full article ">Figure 4
<p>A line plot visualizing the classification accuracy of our method compared to others.</p>
Full article ">Figure 5
<p>A heatmap of the classification accuracy of various methods across subjects and the overall mean.</p>
Full article ">
17 pages, 1963 KiB  
Article
Electrical Bioimpedance-Based Monitoring of Intracochlear Tissue Changes After Cochlear Implantation
by Leanne Sijgers, Marlies Geys, Gunnar Geissler, Patrick Boyle, Alexander Huber and Flurin Pfiffner
Sensors 2024, 24(23), 7570; https://doi.org/10.3390/s24237570 - 27 Nov 2024
Viewed by 713
Abstract
Background: This study examined electrical bioimpedance as a biomarker for intracochlear tissue changes after cochlear implant surgery, comparing monopolar, three-point, and four-point impedance measurements over time and evaluating different measurement systems and approaches. Methods: Impedance measurements were obtained from 21 participants during surgery [...] Read more.
Background: This study examined electrical bioimpedance as a biomarker for intracochlear tissue changes after cochlear implant surgery, comparing monopolar, three-point, and four-point impedance measurements over time and evaluating different measurement systems and approaches. Methods: Impedance measurements were obtained from 21 participants during surgery and at four postoperative stages. Monopolar impedances were recorded using the Bionic Ear Data Collection System (BEDCS) and the Active Insertion Monitoring (AIM) system. Three- and four-point impedances were recorded directly using BEDCS, and indirect three-point impedances were additionally derived from Electrical Field Imaging matrices recorded using BEDCS or AIM. Results: There was an 11% relative error between monopolar measurements from BEDCS and AIM and a 25% discrepancy between direct and indirect three-point measurements. Despite this, direct and indirect measurements from both systems were useful for tracking postoperative impedance shifts. Three- and four-point measurements showed a strong relationship both during and after surgery. Our results suggest that three- and four-point measurements are more specific than monopolar impedances in capturing localized tissue changes. Conclusions: Three- and four-point impedance measurements are potential markers of intracochlear tissue changes over time. While direct three-point impedance measurements offer higher accuracy, indirect measurements provide a feasible alternative for monitoring intracochlear changes in clinical settings lacking the option of direct measurements. Full article
(This article belongs to the Special Issue Bioimpedance Sensors for Medical Monitoring and Diagnosis)
Show Figures

Figure 1

Figure 1
<p>Comparison of the monopolar access resistance (R<sub>a</sub>) measurement conducted using BEDCS and AIM, using color coding to indicate the electrode number. All recordings were obtained during the second postoperative recording session.</p>
Full article ">Figure 2
<p>Bland–Altman plots for the monopolar impedance measurement conducted using BEDCS and AIM (<b>left</b>) and the direct and indirect three-point impedance measurements (<b>right</b>).</p>
Full article ">Figure 3
<p>Comparison of the direct three-point impedance measurement conducted using BEDCS, and the indirect measurement derived from the EFI matrix recorded using BEDCS or AIM. Symbols are used to indicate the recording session, while the measurement system is color-coded. In cases where EFI recordings were conducted using both BEDCS and AIM, the indirect three-point impedances were deduced from the BEDCS recordings.</p>
Full article ">Figure 4
<p>Comparison of monopolar R<sub>a</sub> and three-point impedance measurements. Symbols are used to indicate the recording session, with the upper plots employing color coding to indicate electrode number and the lower plots using color coding to differentiate participants. Note that the three-point impedances are displayed on a logarithmic scale, while the <span class="html-italic">x</span>-axis depicting R<sub>a</sub> is linear, in alignment with model 1. In cases where recordings were obtained using multiple systems or methods, recordings with BEDCS were used instead of recordings made with AIM, and direct three-point impedance measurements were favored over indirect recordings.</p>
Full article ">Figure 5
<p>Visualization of the monopolar and three-point impedance measurements against electrode number for each participant and recording session, indicated using color coding. In cases where recordings were obtained using multiple systems or methods, recordings with BEDCS were used instead of recordings made with AIM, and direct three-point impedance measurements were favored over indirect recordings. Participant 3 had a tip fold-over at electrode 4, while participants 12 and 17 had five and three extracochlear electrodes, respectively. Consequently, specific electrodes were deactivated in these participants’ implant settings, indicated by black dots. For participant 12, the deactivated electrodes had open circuits and are thus not depicted in the figure. Additionally, a few recordings from participant 1 were affected by amplifier saturation and were therefore excluded from the figure.</p>
Full article ">Figure 6
<p>Comparison of three- and four-point impedance measurements, including only direct recordings made with BEDCS. Symbols are used to indicate the recording session, with the upper plots employing color coding to indicate electrode number and the lower plots using color coding to differentiate participants. Both the <span class="html-italic">x</span>- and <span class="html-italic">y</span>-axes use a logarithmic scale, in alignment with model 2.</p>
Full article ">
12 pages, 217 KiB  
Essay
The Architecture of Immortality Through Neuroengineering
by Dany Moussa and Hind Moussa
Philosophies 2024, 9(6), 163; https://doi.org/10.3390/philosophies9060163 - 25 Oct 2024
Viewed by 1072
Abstract
From mobile health and wearables to implantable medical devices and neuroprosthetics, the integration of machines into human biology and cognition is expanding. This paper explores the technological advancements that are pushing the human–machine boundaries further, raising profound questions about identity and existence in [...] Read more.
From mobile health and wearables to implantable medical devices and neuroprosthetics, the integration of machines into human biology and cognition is expanding. This paper explores the technological advancements that are pushing the human–machine boundaries further, raising profound questions about identity and existence in digital realms. The development of robots, androids, and AI–human hybrids promises to augment human capabilities beyond current limits. However, alongside these advancements, significant limitations arise: biological, technical, ethical, and legal. This paper further discusses the existential implications of these technological strides. It addresses the philosophical dimensions of mortality, forgiveness, and the significance of death in a world where technological immortality may be within reach. By addressing these questions, the paper seeks to provide a comprehensive analysis of the potential for these advancements to reshape our understanding of existence and the quest for immortality. Full article
14 pages, 5703 KiB  
Article
A Reconfigurable, Nonlinear, Low-Power, VCO-Based ADC for Neural Recording Applications
by Reza Shokri, Yarallah Koolivand, Omid Shoaei, Daniele D. Caviglia and Orazio Aiello
Sensors 2024, 24(19), 6161; https://doi.org/10.3390/s24196161 - 24 Sep 2024
Viewed by 1128
Abstract
Neural recording systems play a crucial role in comprehending the intricacies of the brain and advancing treatments for neurological disorders. Within these systems, the analog-to-digital converter (ADC) serves as a fundamental component, converting the electrical signals from the brain into digital data that [...] Read more.
Neural recording systems play a crucial role in comprehending the intricacies of the brain and advancing treatments for neurological disorders. Within these systems, the analog-to-digital converter (ADC) serves as a fundamental component, converting the electrical signals from the brain into digital data that can be further processed and analyzed by computing units. This research introduces a novel nonlinear ADC designed specifically for spike sorting in biomedical applications. Employing MOSFET varactors and voltage-controlled oscillators (VCOs), this ADC exploits the nonlinear capacitance properties of MOSFET varactors, achieving a parabolic quantization function that digitizes the noise with low resolution and the spikes with high resolution, effectively suppressing the background noise present in biomedical signals. This research aims to develop a reconfigurable, nonlinear voltage-controlled oscillator (VCO)-based ADC, specifically designed for implantable neural recording systems used in neuroprosthetics and brain–machine interfaces. The proposed design enhances the signal-to-noise ratio and reduces power consumption, making it more efficient for real-time neural data processing. By improving the performance and energy efficiency of these devices, the research contributes to the development of more reliable medical technologies for monitoring and treating neurological disorders. The quantization step of the ADC spans from 44.8 mV in the low-amplitude range to 1.4 mV in the high-amplitude range. The circuit was designed and simulated utilizing a 180 nm CMOS process; however, no physical prototype has been fabricated at this stage. Post-layout simulations confirm the expected performance. Occupying a silicon area is 0.09 mm2. Operating at a sampling frequency of 16 kS/s and a supply voltage of 1 volt, this ADC consumes 62.4 µW. Full article
(This article belongs to the Special Issue CMOS Integrated Circuits for Sensor Applications)
Show Figures

Figure 1

Figure 1
<p>Neural signal illustrating action potentials (APs) and background noise obtained from a rat sample.</p>
Full article ">Figure 2
<p>(<b>a</b>) General VCO-based ADC architecture; (<b>b</b>) schematic of the general VCO units.</p>
Full article ">Figure 3
<p>Block diagram of the proposed architecture.</p>
Full article ">Figure 4
<p>Capacitance of two PMOS transistors, the bulk of which are connected to source and drain (blue line) or VDD (orange line).</p>
Full article ">Figure 5
<p>Schematic of the proposed VCO-based ADC.</p>
Full article ">Figure 6
<p>Schematic of the dynamic sign detection comparator.</p>
Full article ">Figure 7
<p>Schematic of the switches for the calibration part.</p>
Full article ">Figure 8
<p>Layout of the proposed ADC.</p>
Full article ">Figure 9
<p>Nonlinear VCO-based ADC frequency variation vs. input voltage (the orange (blue) waveform represents the post (pre)-layout simulation results).</p>
Full article ">Figure 10
<p>Nonlinear VCO-based ADC transfer curve (the orange (blue) waveform represents the post (pre)-layout simulation results).</p>
Full article ">Figure 11
<p>Monte Carlo simulation results considering mismatch models for all the components (mean = 19.80 MHz, standard deviation = 31.51 kHz).</p>
Full article ">Figure 12
<p>Monte Carlo simulation results considering mismatch models for current mirrors and the common mode feedback amplifier (mean = 19.80 MHz, standard deviation = 30.03 kHz).</p>
Full article ">Figure 13
<p>Simulation results for different DC voltages of the Av1.</p>
Full article ">
6 pages, 1270 KiB  
Perspective
Heterogating Gel Iontronics: A Revolution in Biointerfaces and Ion Signal Transmission
by Zhixin Wu and Ziguang Zhao
Gels 2024, 10(9), 594; https://doi.org/10.3390/gels10090594 - 15 Sep 2024
Viewed by 914
Abstract
Currently, existing iontronic systems are limited and struggle to process electronic-to-multi-ionic transport, resulting in interchange inefficiencies and incompatibilities between artificial ion devices and biological tissue interfaces. The development of heterogating gel iontronics offers a significant advancement in bridging this gap, drawing inspiration from [...] Read more.
Currently, existing iontronic systems are limited and struggle to process electronic-to-multi-ionic transport, resulting in interchange inefficiencies and incompatibilities between artificial ion devices and biological tissue interfaces. The development of heterogating gel iontronics offers a significant advancement in bridging this gap, drawing inspiration from the complex ionic transmission mechanisms found in biological synapses within neural networks. These heterogating gels utilize a biphasic architecture, where the heterointerface effect constructs ionic transfer energy barriers, enabling distinct signal transmission among different ions. In systems with multiple ion species, heterogating gel iontronics allow for precise control of ion transmission, realizing hierarchical and selective cross-stage signal transmission as a neuromorphic function. This perspective highlights the vast potential of heterogating iontronics in applications such as biosensing, neuroprosthetics, and ion separation technologies. Meanwhile, it also addresses the current challenges, including scaling production, ensuring biocompatibility, and integrating with existing technologies, which are crucial for future development. The advancement of heterogating gels is expected to promote the integration between abiotic and biotic systems, with broad implications for smart sensors, bioneural devices, and beyond. Full article
Show Figures

Figure 1

Figure 1
<p>The internal structure of biological neural networks (<b>a</b>) and heterogating gel iontronics (<b>b</b>). The phase separation interface can act as a functional gating that facilitates selective ion transmission. (Note: Numbers represent different ions.)</p>
Full article ">Figure 2
<p>The distinction of the transfer energy barrier between different ions, accompanied by the switching between hydration and partial hydration states across the heterogeneous interfaces, determines the prioritization and hierarchy of multi-ion transmission.</p>
Full article ">Figure 3
<p>Heterogating iontronics with electronic-to-multi-ionic signal transmission demonstrate promising biosensing applications for abiotic–biotic systems.</p>
Full article ">
16 pages, 506 KiB  
Review
The Current Update of Conventional and Innovative Treatment Strategies for Central Nervous System Injury
by Meng-Hsuan Tsai, Chi-Ying Wu, Chao-Hsin Wu and Chun-Yu Chen
Biomedicines 2024, 12(8), 1894; https://doi.org/10.3390/biomedicines12081894 - 19 Aug 2024
Cited by 1 | Viewed by 1525
Abstract
This review explores the complex challenges and advancements in the treatment of traumatic brain injury (TBI) and spinal cord injury (SCI). Traumatic injuries to the central nervous system (CNS) trigger intricate pathophysiological responses, frequently leading to profound and enduring disabilities. This article delves [...] Read more.
This review explores the complex challenges and advancements in the treatment of traumatic brain injury (TBI) and spinal cord injury (SCI). Traumatic injuries to the central nervous system (CNS) trigger intricate pathophysiological responses, frequently leading to profound and enduring disabilities. This article delves into the dual phases of injury—primary impacts and the subsequent secondary biochemical cascades—that worsen initial damage. Conventional treatments have traditionally prioritized immediate stabilization, surgical interventions, and supportive medical care to manage both the primary and secondary damage associated with central nervous system injuries. We explore current surgical and medical management strategies, emphasizing the crucial role of rehabilitation and the promising potential of stem cell therapies and immune modulation. Advances in stem cell therapy, gene editing, and neuroprosthetics are revolutionizing treatment approaches, providing opportunities not just for recovery but also for the regeneration of impaired neural tissues. This review aims to emphasize emerging therapeutic strategies that hold promise for enhancing outcomes and improving the quality of life for affected individuals worldwide. Full article
(This article belongs to the Special Issue Molecular Mechanisms and Novel Therapies for Brain Injury)
Show Figures

Figure 1

Figure 1
<p>Treatment strategies for current approaches and innovative therapies in CNS injuries.</p>
Full article ">
36 pages, 2812 KiB  
Review
Emerging Medical Technologies and Their Use in Bionic Repair and Human Augmentation
by Albert Manero, Viviana Rivera, Qiushi Fu, Jonathan D. Schwartzman, Hannah Prock-Gibbs, Neel Shah, Deep Gandhi, Evan White, Kaitlyn E. Crawford and Melanie J. Coathup
Bioengineering 2024, 11(7), 695; https://doi.org/10.3390/bioengineering11070695 - 9 Jul 2024
Cited by 1 | Viewed by 3455
Abstract
As both the proportion of older people and the length of life increases globally, a rise in age-related degenerative diseases, disability, and prolonged dependency is projected. However, more sophisticated biomedical materials, as well as an improved understanding of human disease, is forecast to [...] Read more.
As both the proportion of older people and the length of life increases globally, a rise in age-related degenerative diseases, disability, and prolonged dependency is projected. However, more sophisticated biomedical materials, as well as an improved understanding of human disease, is forecast to revolutionize the diagnosis and treatment of conditions ranging from osteoarthritis to Alzheimer’s disease as well as impact disease prevention. Another, albeit quieter, revolution is also taking place within society: human augmentation. In this context, humans seek to improve themselves, metamorphosing through self-discipline or more recently, through use of emerging medical technologies, with the goal of transcending aging and mortality. In this review, and in the pursuit of improved medical care following aging, disease, disability, or injury, we first highlight cutting-edge and emerging materials-based neuroprosthetic technologies designed to restore limb or organ function. We highlight the potential for these technologies to be utilized to augment human performance beyond the range of natural performance. We discuss and explore the growing social movement of human augmentation and the idea that it is possible and desirable to use emerging technologies to push the boundaries of what it means to be a healthy human into the realm of superhuman performance and intelligence. This potential future capability is contrasted with limitations in the right-to-repair legislation, which may create challenges for patients. Now is the time for continued discussion of the ethical strategies for research, implementation, and long-term device sustainability or repair. Full article
(This article belongs to the Special Issue Medical Devices and Implants, 2nd Edition)
Show Figures

Graphical abstract

Graphical abstract
Full article ">Figure 1
<p>Brain–machine interface (BMI) design and operation. Electrical or other signals reflecting brain activity are recorded from the scalp, the cortical surface, or within the brain. Magnetoencephalography (MEG) detects magnetic fields created as individual neurons “fire” within the brain, pinpointing the active region within a millimeter and can follow the movement of brain activity as it travels from region to region. Functional magnetic resonance imaging (fMRI) exploits the changes in the magnetic properties of hemoglobin as it carries oxygen. Activation of a part of the brain can increase the ratio of oxyhemoglobin to deoxyhemoglobin. In a similar way to non-invasive EEG, electrocorticography (ECoG) detects and measures the electrical activity of the brain; however, ECoG measurements are taken following direct electrode contact with the cortical surface of the skull. Thus, ECoG has become a tool for detecting brain activity with higher-quality signals to EEG [<a href="#B11-bioengineering-11-00695" class="html-bibr">11</a>]. These signals are analyzed to measure signal features (e.g., single neuron firing rates, amplitudes of EEG rhythms) before their translation into commands that operate applications to replace, restore, enhance, supplement, or improve natural central nervous system outputs [<a href="#B12-bioengineering-11-00695" class="html-bibr">12</a>]. Many commercial ECoG electrode arrays are used clinically and differ in their shape, number of electrodes, spacing, thickness, and materials used. In recent clinical fields, ECoG electrodes are generally implemented for invasive extra-operative monitoring in, for example, patients with drug-resistant epilepsy, and in identifying precise seizure onset zones for resective surgery [<a href="#B13-bioengineering-11-00695" class="html-bibr">13</a>,<a href="#B14-bioengineering-11-00695" class="html-bibr">14</a>]. ECoG’s feasibility is also increasingly being used for rehabilitation purposes in patients with locked-in syndrome, and spinal cord injury [<a href="#B11-bioengineering-11-00695" class="html-bibr">11</a>].</p>
Full article ">Figure 2
<p>Classification of exoskeletons. Recent research has focused on load augmentation for soldiers/workers, assisting trauma patients, paraplegics, spinal cord injured (SCI) persons and for rehabilitation purposes. For medical exoskeletons, the motion trajectories for individual joints cannot be provided by the wearer as the patient cannot make the required movements. Thus, user interfaces, control strategies, mechanical interfaces, etc., need to be designed specifically to cater for the individualistic needs of the patient. For non-medical exoskeleton applications, the methods for measuring “user intention” are most important; therefore, facilitating actuated mechanisms that support the user’s actions and thus ensuring the desired motions are as natural as possible, is the key objective. HAL: Hybrid Assistive Leg, CHRIS: Cybernetic Human–Robot Interface System, BLEEX: Berkeley Lower Extremity Exoskeleton, MIT: Massachusetts Institute of Technology, SCI: spinal cord injury [<a href="#B108-bioengineering-11-00695" class="html-bibr">108</a>,<a href="#B128-bioengineering-11-00695" class="html-bibr">128</a>,<a href="#B129-bioengineering-11-00695" class="html-bibr">129</a>,<a href="#B130-bioengineering-11-00695" class="html-bibr">130</a>,<a href="#B132-bioengineering-11-00695" class="html-bibr">132</a>,<a href="#B133-bioengineering-11-00695" class="html-bibr">133</a>,<a href="#B141-bioengineering-11-00695" class="html-bibr">141</a>,<a href="#B142-bioengineering-11-00695" class="html-bibr">142</a>,<a href="#B143-bioengineering-11-00695" class="html-bibr">143</a>,<a href="#B144-bioengineering-11-00695" class="html-bibr">144</a>,<a href="#B145-bioengineering-11-00695" class="html-bibr">145</a>,<a href="#B146-bioengineering-11-00695" class="html-bibr">146</a>,<a href="#B147-bioengineering-11-00695" class="html-bibr">147</a>,<a href="#B148-bioengineering-11-00695" class="html-bibr">148</a>,<a href="#B149-bioengineering-11-00695" class="html-bibr">149</a>,<a href="#B150-bioengineering-11-00695" class="html-bibr">150</a>,<a href="#B151-bioengineering-11-00695" class="html-bibr">151</a>,<a href="#B152-bioengineering-11-00695" class="html-bibr">152</a>,<a href="#B153-bioengineering-11-00695" class="html-bibr">153</a>,<a href="#B154-bioengineering-11-00695" class="html-bibr">154</a>,<a href="#B155-bioengineering-11-00695" class="html-bibr">155</a>,<a href="#B156-bioengineering-11-00695" class="html-bibr">156</a>,<a href="#B157-bioengineering-11-00695" class="html-bibr">157</a>].</p>
Full article ">Figure 3
<p>(<b>A</b>) The first retinal implant (ARGUS I) was developed in 2002, and later, ARGUS II<sup>®</sup> received FDA approval in 2013. Presently, Retina Implant Alpha IMS<sup>®</sup> has undergone clinical trials; the PRIMA bionic vision system, the IRIS V2 and Suprachoroidal Retinal Prosthesis have also been tested in human studies. The EPI-RET3, Subretinal Retinal Prosthesis and the fully organic P3HT prosthesis are more recent devices that have been studied using pre-clinical models only. (<b>B</b>) The P3HT retinal polymer prosthetic is biocompatible and can cause strong neural responses in the same way as when naturally responding to impulses from rods or cones. This light-sensitive implant material is able to extend the wavelength over which an animal is able to detect light as well as improve visual acuity in rodents. Although proven to be efficient, the working mechanism is not fully understood [<a href="#B198-bioengineering-11-00695" class="html-bibr">198</a>].</p>
Full article ">Figure 4
<p>Biohacking can be as simple as changing your diet, listening to music, and taking supplements, to trying to change your gut microbiome, gene therapy, or methods to modify genetic or brain function to improve one’s self (faster, stronger, mitigating a predisposition for a disease, better focus, memory, energy, etc.) It also refers to devices that may extend or improve human capabilities to enhance the human condition such as exoskeletons and implantable biosensors.</p>
Full article ">Figure 5
<p>The promise of implantable, wearable, and genetic technology for improving human health, wellbeing, and performance has never been greater. However, novel smart technologies that are able to biointerface with the body and/or society, raises social, ethical, and environmental issues. As this field continues to progress and their use becomes increasingly inseparable from the human world, the implementation of strategies that address these issues are needed.</p>
Full article ">
15 pages, 6159 KiB  
Article
High-Porosity Sieve-Type Neural Electrodes for Motor Function Recovery and Nerve Signal Acquisition
by Wonsuk Choi, HyungDal Park, Seonghwan Oh, Seonho Seok, Dae Sung Yoon and Jinseok Kim
Micromachines 2024, 15(7), 862; https://doi.org/10.3390/mi15070862 - 30 Jun 2024
Viewed by 3879
Abstract
In this study, the effects of electrode porosity on nerve regeneration and functional recovery after sciatic nerve transection in rats was investigated. A sieve-type neural electrode with 70% porosity was designed and compared with an electrode with 30% porosity. Electrodes were fabricated from [...] Read more.
In this study, the effects of electrode porosity on nerve regeneration and functional recovery after sciatic nerve transection in rats was investigated. A sieve-type neural electrode with 70% porosity was designed and compared with an electrode with 30% porosity. Electrodes were fabricated from photosensitive polyimide and implanted into the transected sciatic nerves. Motor function recovery was evaluated using the Sciatic Function Index. The number of active channels and their signal quality were recorded and analyzed to assess the sensory neural signal acquisition. Electrical impedance spectroscopy was used to evaluate the electrode performance. The group implanted with the 70% porosity electrode demonstrated significantly enhanced nerve regeneration and motor function recovery, approaching control group levels by the fifth week. In contrast, the group with the 30% porosity electrode exhibited limited improvement. Immunohistochemical analysis confirmed extensive nerve fiber growth within the 70% porous structure. Moreover, the 70% porosity electrode consistently acquired neural signals from more channels compared to the 30% porosity electrode, demonstrating its superior performance in sensory signal detection. These findings emphasize the importance of optimizing electrode porosity in the development of advanced neural interfaces, with the potential to enhance clinical outcomes in peripheral nerve repair and neuroprosthetic applications. Full article
(This article belongs to the Special Issue Neural Interface: From Material to System)
Show Figures

Figure 1

Figure 1
<p>Porosity and microelectrode channel count of various sieve-type neural electrodes: This summarizes the porosity and microelectrode channel count of various sieve-type neural electrodes reported in prior studies. The table details specific designs, while the graph visualizes these data, with the blue circle representing the electrode proposed in this study [<a href="#B33-micromachines-15-00862" class="html-bibr">33</a>,<a href="#B34-micromachines-15-00862" class="html-bibr">34</a>,<a href="#B35-micromachines-15-00862" class="html-bibr">35</a>,<a href="#B36-micromachines-15-00862" class="html-bibr">36</a>,<a href="#B37-micromachines-15-00862" class="html-bibr">37</a>].</p>
Full article ">Figure 2
<p>Overview of the high-porosity, sieve-type neural interface: (<b>A</b>) Schematic illustration of a rat with a high-porosity sieve-type neural electrode implanted in the sciatic nerve, connected to the recording system. (<b>B</b>) Magnified view of the neural electrode, highlighting the arrangement of the regenerative nerve fibers within the interface.</p>
Full article ">Figure 3
<p>Design and layer composition of the high-porosity sieve-type neural electrode: (<b>A</b>) Sieve design over the 70% porosity area, which supports nerve regeneration. The internal arrangement includes a ring-shaped configuration with 32 working electrodes (gold) and three reference elec-trodes (black arrows). (<b>B</b>) Electrode’s layered structure, featuring photosensitive polyimide (PSPI), and metal layers of titanium/gold/titanium.</p>
Full article ">Figure 4
<p>Visualization of sieve-type neural electrodes. (<b>A</b>) Optical optical images displaying sieve-type neural electrodes with 70% porosity (<b>up</b>) and 30% porosity (<b>down</b>). The scale bar is 600 µm. (<b>B</b>) Scanning electron microscope (SEM) images of the 70% porosity sieve-type neural electrode, showing the regenerative ring-shaped arrangement of the 32 working and reference electrodes. The left image scale bar is 500 µm. The right image is a magnified view of a working electrode with a scale bar of 50 µm. (<b>C</b>) SEM images detailing the structure of the 30% porosity sieve-type neural electrode. The left image scale bar is 500 µm. The right image is a magnified view of a working electrode with a scale bar of 50 µm.</p>
Full article ">Figure 5
<p>Implantation surgery of the rat sciatic nerve. (<b>A</b>) The top image shows the 70% porosity sieve-type neural electrode before implantation. The bottom image displays the electrode after sci-atic nerve transection and insertion, secured with four suture points. The scale bars are 2 mm (<b>top</b>) and 2.5 mm (<b>bottom</b>). (<b>B</b>) The top image illustrates the 30% porosity sieve-type neural electrode before implantation. The bottom image shows the electrode inserted into the transected sciatic nerve. The scale bars are 1.5 mm (<b>top</b>) and 2 mm (<b>bottom</b>).</p>
Full article ">Figure 6
<p>Electrical characteristics of sieve-type neural electrodes. (<b>A</b>) Graph showing the impedance and phase characteristics of the 70% porosity sieve-type neural electrode (blue square boxes and solid lines) and the 30% porosity sieve-type neural electrode (black circles and solid lines). (<b>B</b>) Graph showing the impedance recorded at a frequency of 1 kHz over an 8-week period following surgical implantation into the sciatic nerve for the 70% porosity sieve-type neural electrode (blue square boxes and solid lines) and the 30% porosity sieve-type neural electrode (black circles and dashed lines).</p>
Full article ">Figure 7
<p>Sciatic functional index (SFI) and footprint analysis for motor function recovery. The left (<b>A</b>) graph displays the SFI over an 8-week period for the transection model (black), 30% porosity sieve-type neural electrode (light blue), and 70% porosity sieve-type neural electrode (dark blue). Error bars indicate standard deviations. The right panels (<b>B</b>) show representative footprint patterns at 1, 2, 4, and 8 weeks post-surgery. The experimental groups (transection, 70% porosity sieve, and 30% porosity sieve) are compared to normal controls. The 8-week footprints include detailed markers for paw length (PL), toe spread (TS), and intermediary toe spread (IT), illustrating the extent of functional recovery.</p>
Full article ">Figure 8
<p>Sensory Signal Acquisition and Analysis. (<b>A</b>) Representative recordings of raw and filtered sensory signals obtained from the 70% porosity sieve-type neural electrode. These panels show time-domain signals, with the top graph displaying the raw signal in blue and the bottom graph showing the filtered signal in red. (<b>B</b>) This graph presents a multi-dimensional analysis, including the principal component analysis (PCA) plot, showing the differentiation between signal clusters in (<b>C</b>). This data is recorded from a single working electrode channel at the 2-week mark. (<b>D</b>) The graph depicts the raw data and filtered data in the time domain, with the fast Fourier transform (FFT) results on the bottom left. The graphs display averaged sensory signal waveforms and their corresponding standard deviations in (<b>E</b>) positive and (<b>F</b>) negative action potentials. This data is also from a single working electrode channel at the 2-week mark. (<b>G</b>) Bar graph comparing the number of sensory signal-acquired channels between the 30% porosity model (light blue) and the 70% porosity model (dark blue) over an 8-week period. Error bars represent standard deviations. Each bar represents the average number of active channels and their standard deviations, demonstrating that the 70% porosity model consistently achieves higher signal acquisition, indicating its superior performance in sensory signal detection.</p>
Full article ">Figure 9
<p>Immunohistochemistry (IHC) analysis post-implantation of the 70% porosity sieve-type neural electrode. Differential interference contrast (DIC) image (<b>left</b>), DAPI staining for cell nuclei in blue (<b>middle</b>), and neurofilament staining (NFM) in green (<b>right</b>) are shown for the sciatic nerve 8 weeks after implantation of the 70% porosity sieve-type neural electrode. The images illustrate the integration of the neural electrode with the surrounding nerve tissue, highlighting the structural integrity and cellular organization around the implant. Scale bar represents 100 µm.</p>
Full article ">
15 pages, 5193 KiB  
Article
Wavelet Transforms Significantly Sparsify and Compress Tactile Interactions
by Ariel Slepyan, Michael Zakariaie, Trac Tran and Nitish Thakor
Sensors 2024, 24(13), 4243; https://doi.org/10.3390/s24134243 - 29 Jun 2024
Viewed by 1311
Abstract
As higher spatiotemporal resolution tactile sensing systems are being developed for prosthetics, wearables, and other biomedical applications, they demand faster sampling rates and generate larger data streams. Sparsifying transformations can alleviate these requirements by enabling compressive sampling and efficient data storage through compression. [...] Read more.
As higher spatiotemporal resolution tactile sensing systems are being developed for prosthetics, wearables, and other biomedical applications, they demand faster sampling rates and generate larger data streams. Sparsifying transformations can alleviate these requirements by enabling compressive sampling and efficient data storage through compression. However, research on the best sparsifying transforms for tactile interactions is lagging. In this work we construct a library of orthogonal and biorthogonal wavelet transforms as sparsifying transforms for tactile interactions and compare their tradeoffs in compression and sparsity. We tested the sparsifying transforms on a publicly available high-density tactile object grasping dataset (548 sensor tactile glove, grasping 26 objects). In addition, we investigated which dimension wavelet transform—1D, 2D, or 3D—would best compress these tactile interactions. Our results show that wavelet transforms are highly efficient at compressing tactile data and can lead to very sparse and compact tactile representations. Additionally, our results show that 1D transforms achieve the sparsest representations, followed by 3D, and lastly 2D. Overall, the best wavelet for coarse approximation is Symlets 4 evaluated temporally which can sparsify to 0.5% sparsity and compress 10-bit tactile data to an average of 0.04 bits per pixel. Future studies can leverage the results of this paper to assist in the compressive sampling of large tactile arrays and free up computational resources for real-time processing on computationally constrained mobile platforms like neuroprosthetics. Full article
Show Figures

Figure 1

Figure 1
<p>High-density tactile compression in a robotic hand. (1) High-density tactile sensors produce dense spatiotemporal tactile data. (2) Local compression is employed to significantly reduce the bandwidth for data transmission of the robotic hand. (3) Tactile data are reconstructed from the measured compressed outputs.</p>
Full article ">Figure 2
<p>Overview of our wavelet transform-based sparsification approach applied to tactile compression scheme in the 2D case. The sparsification scheme is laid out as follows: tactile data are collected from the sensor array and a wavelet transform is performed on each collected frame. The values after transformation are quantized and can now be transmitted or stored. To recover the compressed data, they must be unquantized and the inverse wavelet transform should be applied. The reconstructed data will have small deviations from the original data due to losses during quantization. During the survey, metrics such as the sparsity, average number of bits per pixel, energy ratio, and normalized mean square error are calculated. The tactile sensor array presented has 548 sensors arranged in a 32 × 32 matrix, and lighter colors represent higher pressures.</p>
Full article ">Figure 3
<p>Three-level discrete wavelet transform implemented as cascading filterbanks with downsampling. LoD represents the low-pass filter and HiD represents the high-pass filter. ↓2 represents downsampling by 2. x[n] is the input signal. ai[n] is the approximation coefficients at level i. di[n] are the detail coefficients at level i. The output length is equivalent to the input length.</p>
Full article ">Figure 4
<p>Twelve wavelet families used in this survey. An example of the wavelet function is shown for each. The number of iterations used is 5. Discrete Meyer (dmey) is zoomed in on the x-axis for clarity.</p>
Full article ">Figure 5
<p>Sparsification and compression by quantization. Quantization is performed with Equation (1) to achieve a desired NMSE = 0.01. The left side shows the original signal and its unquantized wavelet representations. The right side shows the reconstructed signal and the quantized wavelet representations. Small magnitude coefficients are quantized to zero, generating sparsity in the representation.</p>
Full article ">Figure 6
<p>Compactness of 1D wavelet representation of tactile data of an exemplary grasping event. The effect of number of coefficients on NMSE and PSNR, and their ranking order for an exemplary grasping event is shown. A legend is shown at the bottom of the figure to associate the lines in the plots with their matching transform. The best-performing 1D wavelet transforms from a few families are highlighted.</p>
Full article ">Figure 7
<p>Sparsity of candidate transforms and dimensionality trend. Each candidate transform is represented as a marker with a particular shape, edge color, and face color that encodes its name and size using the legend shown at the bottom of the figure. Colored regions represent the distribution of sparsity for each dimensional transform. Red, green, and blue represent 1D, 2D, and 3D transforms, respectively. Each region is centered at the mean value (darker line) and its vertical span shows the standard deviation of sparsity for that dimensional transform.</p>
Full article ">Figure 8
<p>Sparsity and compressibility of the best candidate wavelet transforms for each dimensionality case versus NMSE in the left. The sparsity of each dimensionality case in aggregate is presented in the right.</p>
Full article ">Figure 9
<p>Temporal and spatial errors of tactile reconstruction for the sparsest wavelet transforms in each dimensionality group. The temporal error is calculated as the error over time for each sensor independently and the spatial error is calculated as the error at each spatial frame for each time point independently. All 3 reconstructions have the same average NMSE error (0.01), but different spatial and temporal standard deviations depending on the dimensionality of the transform used. Solid lines represent the average error in each graph, and colored zones represent the span of the standard deviation.</p>
Full article ">Figure 10
<p>Visual effect of transform dimensionality on tactile reconstruction of a single selected grasp. Top row shows the tactile data over time for each sensor in the array during the grasp. The original curves are shown on the left, and the reconstructed curves for each dimensional transform are shown in the following graphs with each having an overall equivalent NMSE of 0.01. The red dashed line in each graph shows the selected frame during the grasp for which the pressure over the array is plotted spatially. The frame error of each reconstructed frame is emphasized to help discriminate the spatial differences. Respectively, 1D has the highest spatial error of 0.0196, 2D has the lowest spatial error of 0.0151, and 3D has the middle spatial error of 0.0167.</p>
Full article ">Figure 11
<p>Relationship between filter size and sparsity for different NMSE thresholds for 1D transforms. Markers are coded using the same legend as <a href="#sensors-24-04243-f007" class="html-fig">Figure 7</a>. The trend line is shown in yellow as a linear fit of the data, with a dashed pink line showing the associated 95% confidence interval. Relationship between filter size and sparsity changes from smaller filters being most sparsifying for coarse estimates (NMSE = 0.01), to having no relationship at NMSE = 0.0043, to having larger filters be more sparsifying for high-accuracy reconstructions (NMSE = 0.0001).</p>
Full article ">Figure 12
<p>The most sparsifying candidate wavelets for each NMSE and their similarity to the grand average tactile interaction. The grand average tactile interaction is shown on the left, with labeled standard deviation. The scaling and wavelet function of the sparsest candidate for each NMSE value are plotted below the respective section. They are labelled with their corresponding color-coded markers for comparison to other figures. These best wavelets are all in the 1D case: Bior6.8, Bior4.4, Bior4.4, and Sym4 for NMSEs of 0.0001, 0.0015, 0.0043, and 0.01, respectively.</p>
Full article ">
2 pages, 158 KiB  
Editorial
Neuroprosthetics of the Hand: Current Hot Research Topics, Research Trends and Challenges, and Recent Innovations
by Andreas Otte
Prosthesis 2024, 6(3), 670-671; https://doi.org/10.3390/prosthesis6030047 - 12 Jun 2024
Viewed by 1049
Abstract
If you want to buy a new car today, you can expect a flood of digital features and assistance systems that initially make the analog human heart beat faster [...] Full article
(This article belongs to the Special Issue Prosthesis: Spotlighting the Work of the Editorial Board Members)
23 pages, 13655 KiB  
Article
Prediction of Hippocampal Signals in Mice Using a Deep Learning Approach for Neurohybrid Technology Applications
by Albina V. Lebedeva, Margarita I. Samburova, Vyacheslav V. Razin, Nikolay V. Gromov, Svetlana A. Gerasimova, Tatiana A. Levanova, Lev A. Smirnov and Alexander N. Pisarchik
Algorithms 2024, 17(6), 252; https://doi.org/10.3390/a17060252 - 7 Jun 2024
Viewed by 1406
Abstract
The increasing growth in knowledge about the functioning of the nervous system of mammals and humans, as well as the significant neuromorphic technology developments in recent decades, has led to the emergence of a large number of brain–computer interfaces and neuroprosthetics for regenerative [...] Read more.
The increasing growth in knowledge about the functioning of the nervous system of mammals and humans, as well as the significant neuromorphic technology developments in recent decades, has led to the emergence of a large number of brain–computer interfaces and neuroprosthetics for regenerative medicine tasks. Neurotechnologies have traditionally been developed for therapeutic purposes to help or replace motor, sensory or cognitive abilities damaged by injury or disease. They also have significant potential for memory enhancement. However, there are still no fully developed neurotechnologies and neural interfaces capable of restoring or expanding cognitive functions, in particular memory, in mammals or humans. In this regard, the search for new technologies in the field of the restoration of cognitive functions is an urgent task of modern neurophysiology, neurotechnology and artificial intelligence. The hippocampus is an important brain structure connected to memory and information processing in the brain. The aim of this paper is to propose an approach based on deep neural networks for the prediction of hippocampal signals in the CA1 region based on received biological input in the CA3 region. We compare the results of prediction for two widely used deep architectures: reservoir computing (RC) and long short-term memory (LSTM) networks. The proposed study can be viewed as a first step in the complex task of the development of a neurohybrid chip, which allows one to restore memory functions in the damaged rodent hippocampus. Full article
Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Scheme of the experiment on recording field excitatory postsynaptic potentials (fEPSPs) in mice hippocampal slices and scheme for training neural networks (LSTM or reservoir). The left panel shows the mouse hippocampal slices with a protocol for installing, recording and stimulating electrodes for electrical stimulation and subsequent synaptic transmission activation in the perforant (3-synaptic) pathway of the hippocampus. A stimulating electrode was placed in the DG area and sent electrical square current pulses to activate the cells in DG area of hippocampus. Recording electrodes were installed in the pyramidal neuron dendrites of CA3 and CA1, the hippocampus areas. This made it possible to record the activation of the perforant pathway in the hippocampus due to electrical stimulation. The following shows the original trace of the fEPSP recorded in the dendrites of CA3 and CA1 hippocampus areas. These signals were fed to the input of LSTM or reservoir for training. The right panel shows the architectures of the neural networks used. After training these neural networks with fEPSP signals recorded from the CA3 region in hippocampal slices (input signals), predicted signals for the CA1 region (output signals) were obtained. (<b>b</b>) Representative examples of original fEPSP traces at 400 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A and 500 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitudes.</p>
Full article ">Figure 2
<p>Pipeline for data processing which includes four main steps.</p>
Full article ">Figure 3
<p>LSTM architecture used for prediction of fEPSP signal prediction in CA1 region using CA3 signal as an input.</p>
Full article ">Figure 4
<p>Reservoir architecture used for fEPSP signal prediction in CA1 region using CA3 signal as an input.</p>
Full article ">Figure 5
<p>Signal parameters for subsequent evaluation of the predicted signal quality metrics: 1—rise time, 2—decrease time, 3—response time halfwidth, 4—amplitude, 5—slope.</p>
Full article ">Figure 6
<p>Boxplots for main features used in custom metric of CA3 signals (left panel) and CA1 signals (right panel) obtained as a response to short rectangular electrical pulse of varying amplitude. (<b>a</b>) Rise time, (<b>b</b>) decrease time, (<b>c</b>) halfwidth, (<b>d</b>) amplitude, (<b>e</b>) slope. Black dots are outliers.</p>
Full article ">Figure 7
<p>Typical examples of signals belonging to different classes. (<b>a</b>) Signals belonging to Class 1, (<b>b</b>) signals belonging to Class 2, (<b>c</b>) signals belonging to Class 3. See detailed description in the text.</p>
Full article ">Figure 8
<p>True (blue) and predicted (red and black) fEPSP signals at 400 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude for (<b>a</b>) LSTM (red) and (<b>b</b>) RC (black) networks.</p>
Full article ">Figure 9
<p>Evaluation metrics for predicted fEPSP signals. (<b>a</b>) MAPE, (<b>b</b>) custom metric based on valuable properties of biological signal.</p>
Full article ">Figure 10
<p>Comparison of prediction quality for LSTM (left panels, red dots) and RC (right panels, black dots) for different custom metric parameters: (<b>a</b>) rise time, (<b>b</b>) decrease time, (<b>c</b>) halfwidth, (<b>d</b>) amplitude and (<b>e</b>) slope.</p>
Full article ">Figure A1
<p>True and predicted fEPSP signals at 100 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A2
<p>True and predicted fEPSP signals at 200 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A3
<p>True and predicted fEPSP signals at 300 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A4
<p>True and predicted fEPSP signals at 500 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A5
<p>True and predicted fEPSP signals at 600 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A6
<p>True and predicted fEPSP signals at 700 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A7
<p>True and predicted fEPSP signals at 800 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A8
<p>True and predicted fEPSP signals at 900 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">Figure A9
<p>True and predicted fEPSP signals at 1000 <math display="inline"><semantics> <mi mathvariant="sans-serif">μ</mi> </semantics></math>A stimulus amplitude. Blue marker corresponds to true CA1 signal, red marker to LSTM-predicted signal, black marker to RC-predicted signal.</p>
Full article ">
12 pages, 2421 KiB  
Article
Flexible Peripheral Nerve Interfacing Electrode for Joint Position Control in Closed-Loop Neuromuscular Stimulation
by Sia Kim and Kang-Il Song
Micromachines 2024, 15(5), 594; https://doi.org/10.3390/mi15050594 - 29 Apr 2024
Viewed by 1239
Abstract
Addressing peripheral nerve disorders with electronic medicine poses significant challenges, especially in replicating the dynamic mechanical properties of nerves and understanding their functionality. In the field of electronic medicine, it is crucial to design a system that thoroughly understands the functions of the [...] Read more.
Addressing peripheral nerve disorders with electronic medicine poses significant challenges, especially in replicating the dynamic mechanical properties of nerves and understanding their functionality. In the field of electronic medicine, it is crucial to design a system that thoroughly understands the functions of the nervous system and ensures a stable interface with nervous tissue, facilitating autonomous neural adaptation. Herein, we present a novel neural interface platform that modulates the peripheral nervous system using flexible nerve electrodes and advanced neuromodulation techniques. Specifically, we have developed a surface-based inverse recruitment model for effective joint position control via direct electrical nerve stimulation. Utilizing barycentric coordinates, this model constructs a three-dimensional framework that accurately interpolates inverse isometric recruitment values across various joint positions, thereby enhancing control stability during stimulation. Experimental results from rabbit ankle joint control trials demonstrate our model’s effectiveness. In combination with a proportional–integral–derivative (PID) controller, it shows superior performance by achieving reduced settling time (less than 1.63 s), faster rising time (less than 0.39 s), and smaller steady-state error (less than 3 degrees) compared to the legacy model. Moreover, the model’s compatibility with recent advances in flexible interfacing technologies and its integration into a closed-loop controlled functional neuromuscular stimulation (FNS) system highlight its potential for precise neuroprosthetic applications in joint position control. This approach marks a significant advancement in the management of neurological disorders with advanced neuroprosthetic solutions. Full article
(This article belongs to the Special Issue Neural Interface: From Material to System)
Show Figures

Figure 1

Figure 1
<p>Animal experimental setup for three-dimensional surface-based inversed recruitment modelling (<b>a</b>) Schematic of animal experimental setup for measuring muscle force while electrical stimulation each on tibial and peroneal nerve at neutral joint position (100°), which is denoted dot lines. And the arrow line denoted the flow of the system. (<b>b</b>) Photograph of direct nerve stimulation electrode.</p>
Full article ">Figure 2
<p>Joint position control experiment using proposed three-dimensional recruitment model. (<b>a</b>) Schematics of joint position control for evaluating the performance of three-dimensional recruitment model. (<b>b</b>) Block diagram of surface-based inverse recruitment model applied on the closed-loop FNS system control.</p>
Full article ">Figure 3
<p>The inverse recruitment curves at different initial joint position (70°, 80°, 90°, 100°, 110°, 120°, and 130°), respectively: inverse recruitment for (<b>a</b>) tibial nerve stimulation, (<b>b</b>) peroneal nerve stimulation. And (<b>c</b>) three-dimensional decision planes obtained for full range of motion, tibial nerve stimulation, and peroneal nerve stimulation.</p>
Full article ">Figure 4
<p>Representative results of ankle position control in plantarflexion at 120°: steady-state responses of (<b>a</b>) surface-based inverse recruitment model and (<b>b</b>) inverse isometric recruitment curve, and corresponding controller trajectories output during ankle-position control in (<b>c</b>) surface-based inverse recruitment model and (<b>d</b>) inverse isometric recruitment curve.</p>
Full article ">Figure 5
<p>Representative results of ankle position control in dorsiflexion at 80°: steady-state responses of (<b>a</b>) surface-based inverse recruitment model and (<b>b</b>) inverse isometric recruitment curve, and corresponding controller trajectories output during ankle-position control in (<b>c</b>) surface-based inverse recruitment model and (<b>d</b>) inverse isometric recruitment curve.</p>
Full article ">Figure 6
<p>Mean absolute errors of ankle-position control: (<b>a</b>) mean absolute errors were evaluated by different inverse recruitment methods at different ankle positions (70°, 90°, 110°, and 130°), and (<b>b</b>) average of mean absolute error in every joint angle position at inverse recruitment and surface inverse recruitment curve, respectively. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test (NS (not significant) <span class="html-italic">p</span> &gt; 0.05).</p>
Full article ">
16 pages, 2936 KiB  
Article
CMOS Analogue Velocity-Selective Neural Processing System
by Shamin Sadrafshari, Sebastian Simmich, Benjamin Metcalfe, Jon Prager, Nicolas Granger, Nick Donaldson, Robert Rieger and John Taylor
Electronics 2024, 13(3), 569; https://doi.org/10.3390/electronics13030569 - 31 Jan 2024
Viewed by 896
Abstract
Velocity-selective recording (VSR) of electroneurogram (ENG) signals is a frequently utilized technology in the field of neural recording with applications in clinical medicine and neuroprosthetics. VSR classifies excited axon populations in terms of their conduction velocities using multiple recordings of the same ENG [...] Read more.
Velocity-selective recording (VSR) of electroneurogram (ENG) signals is a frequently utilized technology in the field of neural recording with applications in clinical medicine and neuroprosthetics. VSR classifies excited axon populations in terms of their conduction velocities using multiple recordings of the same ENG signal and addition of the recording channels after introducing controlled time delays. This paper describes the first fully integrated analogue realization of the complete delay-and-add process with nine channels. The proposed approach uses switched-capacitor (SC) circuits and avoids the need for ADCs at the inputs of the delay-and-add circuit to achieve a small size and low power implementation. Simulated and measured results obtained from chips fabricated in 0.35 µm CMOS technology are reported. The system occupies a 1.16 mm2 active area and consumes 798 µW from a 3 V supply, while achieving a wide velocity detection range of 10–300 m/s with a precise relative velocity resolution down to 0.003. Intrinsic velocity spectra measured from synthetic ENG inputs confirm the operation of the system. Full article
(This article belongs to the Section Bioelectronics)
Show Figures

Figure 1

Figure 1
<p>Schematic of a multi-channel VSR system for recording ENG signals from a 10-electrode cuff, configured as 9 dipole channels (i.e., <span class="html-italic">N</span> = 9). This paper focuses on the analogue VSR delay-and-add circuit (striped pattern) and does not consider either the pre-amplifiers or the dipole amplifiers since they have been previously reported elsewhere.</p>
Full article ">Figure 2
<p>(<b>a</b>) Schematic of an SC realization of a delay-and-add system with two differential inputs (<span class="html-italic">N</span> = 1). (<b>b</b>) Timing diagram for the system. The inputs are sampled onto the capacitors <span class="html-italic">C<sub>i</sub></span> for ultimate accumulation on the capacitors <span class="html-italic">C<sub>f</sub></span>.</p>
Full article ">Figure 3
<p>Timing diagram of the nine-channel delay-and-add circuit. One complete sampling cycle is shown. Note that the output is only available once in each ten cycles, when all the samples have been accumulated on <span class="html-italic">C<sub>f</sub></span> (indicated in green).</p>
Full article ">Figure 4
<p>Schematic of an SC realization of a delay-and-add system with nine inputs (<span class="html-italic">N</span> = 8), an extension of the circuit shown in <a href="#electronics-13-00569-f002" class="html-fig">Figure 2</a> (with <span class="html-italic">N</span> = 1). In order to increase the sampling rate and to meet the bandwidth requirements of the analogue input signals, 10 blocks of time-interleaved input sampling stages have been added in parallel.</p>
Full article ">Figure 5
<p>Timing diagram for a time-interleaved nine-channel VSR system. All the clock signals (<span class="html-italic">φ</span><sub>1–10,<span class="html-italic">x</span></sub>) are generated by ten non-overlapping pulses of width <span class="html-italic">T<sub>d</sub></span> and period 10<span class="html-italic">T<sub>d</sub></span>, delayed by <span class="html-italic">T<sub>d</sub></span> from each other.</p>
Full article ">Figure 6
<p>Timing diagram of the control phases for the first time-interleaved input sampling stage block, shown in <a href="#electronics-13-00569-f004" class="html-fig">Figure 4</a>.</p>
Full article ">Figure 7
<p>Control signal generator circuit. This consists of a synchronous ring counter using D-type flip-flops and some logic.</p>
Full article ">Figure 8
<p>Amplifier circuits and the corresponding component dimensions. The SC amplifier on the left is a folded-cascode design. The transconductance amplifier on the right is the CM stabilisation circuit.</p>
Full article ">Figure 9
<p>Nine-channel VSR system time domain simulation. (<b>a</b>) Time-delayed (10 µs) inputs of channels 1 and 9 and the corresponding sampled output. (<b>b</b>) Timing diagram of the on-chip clock generator circuit for an external clock signal of 100 kHz, representing 10 µs inter-channel delay.</p>
Full article ">Figure 10
<p>Microphotograph of the implemented nine-channel VSR system without the pre-amplifier stage.</p>
Full article ">Figure 11
<p>Measured signals for a sinusoidal input with a frequency of 3 kHz, the sampled output and a subsampled interpolated output signal.</p>
Full article ">Figure 12
<p>Measured Lissajous traces of channel 1 (90 µs delay) and channel 9 (10 µs delay) in comparison with the ideal delay (dashed ellipses).</p>
Full article ">Figure 13
<p>Measured output voltage peak to peak for a sinusoidal input signal. (<b>a</b>) Signal frequency of 5 kHz with different system delays. (<b>b</b>) Different input signal frequencies with the same inter-channel delay of 10 µs.</p>
Full article ">Figure 14
<p>(<b>a</b>) Artificially generated action potentials with an amplitude of 18 mV and a delay of 10 µs between adjacent channels as input to the nine-channel system. (<b>b</b>) Sampled and interpolated output of the system with a matched clock.</p>
Full article ">Figure 15
<p>Two synthetic action potentials with the same amplitude and different velocities—AP1 with 45 m/s (red dashed line) and AP2 with 85 m/s (green dotted–dashed line). Both action potentials were then superimposed to yield the test signals (blue line).</p>
Full article ">Figure 16
<p>Measured intrinsic velocity spectrum (<b>a</b>) and the 8 kHz fourth-order bandpass-filtered intrinsic velocity spectrum (<b>b</b>) for a combination of two artificially generated action potentials with velocities of 45 m/s and 85 m/s. The orange lines are shape-preserving fits through the measured data points.</p>
Full article ">Figure 17
<p>Measured intrinsic velocity spectrum (<b>a</b>) and the 8 kHz fourth-order bandpass-filtered intrinsic velocity spectrum (<b>b</b>) for a combination of two artificially generated action potentials with velocities of 25 m/s and 35 m/s. The orange lines are shape-preserving fits through the measured data points.</p>
Full article ">
20 pages, 6294 KiB  
Article
Neuromorphic Analog Machine Vision Enabled by Nanoelectronic Memristive Devices
by Sergey Shchanikov, Ilya Bordanov, Alexey Kucherik, Evgeny Gryaznov and Alexey Mikhaylov
Appl. Sci. 2023, 13(24), 13309; https://doi.org/10.3390/app132413309 - 16 Dec 2023
Viewed by 1744
Abstract
Arrays of memristive devices coupled with photosensors can be used for capturing and processing visual information, thereby realizing the concept of “in-sensor computing”. This is a promising concept associated with the development of compact and low-power machine vision devices, which is crucial important [...] Read more.
Arrays of memristive devices coupled with photosensors can be used for capturing and processing visual information, thereby realizing the concept of “in-sensor computing”. This is a promising concept associated with the development of compact and low-power machine vision devices, which is crucial important for bionic prostheses of eyes, on-board image recognition systems for unmanned vehicles, computer vision in robotics, etc. This concept can be applied for the creation of a memristor based neuromorphic analog machine vision systems, and here, we propose a new architecture for these systems in which captured visual data are fed to a spiking artificial neural network (SNN) based on memristive devices without analog-to-digital and digital-to-analog conversions. Such an approach opens up the opportunities of creating more compact, energy-efficient visual processing units for wearable, on-board, and embedded electronics for such areas as robotics, the Internet of Things, and neuroprosthetics, as well as other practical applications in the field of artificial intelligence. Full article
(This article belongs to the Special Issue Artificial Intelligence (AI) in Neuroscience)
Show Figures

Figure 1

Figure 1
<p>The main features of the neuromorphic analog machine vision systems, combining “in-memory computing”, “in-sensor-computing”, and neuromorphic architectures. (<b>A</b>) In comparison with the existing digital computer vision systems, “in-memory computing” makes it possible to process visual information entirely within the hardware when the ANN models work completely in analog form using computers based on memristive devices. (<b>B</b>) In comparison with the existing general memristor-based computers, in the neuromorphic analog machine vision systems, there are no analog-to-digital and digital-to-analog conversions in the process of capturing visual information via photosensors. For these purposes, the devices for “in-sensor computation” can be used in the sensory part of a system for capturing visual information in analog form, which is then fed to an ANN based on memristors.</p>
Full article ">Figure 2
<p>Spiking ANNs based on memristive devices. (<b>A</b>) Common architecture includes presynaptic neurons and postsynaptic neurons, connected by artificial synapses implemented with memristors. Presynaptic neurons generate spikes encoding input information. Spikes go through the memristive synapses and locally change their resistances in accordance with the STDP (spike timing-dependent plasticity) rule, providing the self-learning of the whole ANN. (<b>B</b>) The dependence of the change ΔW in synaptic conductance on the interval Δ<span class="html-italic">t</span> between a presynaptic spike and a postsynaptic spike for different current synaptic conductance values W. (<b>C</b>) A spiking neuron receives sequences of spikes on its inputs and, under certain conditions, generates a spike at its output; for example, in the LIF (leaky integrate-and-fire) model, each spike contributes to the neuron’s status—its amplitude, which decays over time; if a sufficient number of spikes contributes to the status in a certain time window, the neuron’s amplitude exceeds a threshold, and the neuron generates an output spike [<a href="#B2-applsci-13-13309" class="html-bibr">2</a>].</p>
Full article ">Figure 3
<p>The electrical circuits of the sensory part (input channels) for the capturing and encoding of visual information in analog form and its subsequent transmission to an ANN without digitization. (<b>A</b>) An input channel for the formal ANNs. It consists of a photodiode PD, a load resistor R<sub>load</sub> (for converting photocurrent i<sub>ph</sub> to voltage), an operational amplifier U<sub>1</sub>, and a bias voltage source V<sub>bias</sub> (reverse biasing is used in the photoconductive mode providing a wider bandwidth, higher sensitivity, and improved linearity (for sensitivity control at different weather conditions, day time, etc.), but also increases noise and dark current). The SPDT and SPST switches control the operating modes of the circuit: the “write mode”, when recording the ANN weights (by changing the memristor’s resistance from R<sub>init</sub> to the target R<sub>T</sub>), and the “inference mode”. Visual information in this case is encoded using the voltage amplitude V<sub>i</sub> in each input channel. (<b>B</b>) An input channel for the spiking ANN. It consists of the same elements as the input channel for the formal ANN, except for the integrator with threshold U<sub>2</sub>. The integrator accumulates charge and generates pulses of the same amplitude but at different frequencies, like an integrate-and-fire neuron (I&amp;F). The multiplexer (MUX) U<sub>3</sub> and the SPDT switch control the operating modes of the circuit: for recording ANN weights, self-learning (by changing the memristor’s resistance from R<sub>init</sub> to R<sub>inf</sub>), and inference. Visual information is encoded according to signal frequency f<sub>i</sub>.</p>
Full article ">Figure 4
<p>The most common options for implementing a memristor-based ANN synapse for the neuromorphic analog computer vision systems. (<b>A</b>) The unipolar weight is formed with one memristor and can be mathematically calculated in different ways: for Equations (1) and (2), the weight plot is a hyperbola, and for Equation (3), it is a straight line from 0 to 1. (<b>B</b>) The bipolar weight is formed by a pair of memristors. The sign is obtained due to the differential output. (<b>C</b>) A similar circuit, but where the sign is obtained due to the differential input. One input x corresponds to two inputs V<sub>in</sub> and −V<sub>in</sub>. The weight plot for Equations (4) and (5) is a hyperbola. (<b>D</b>) The bipolar weight is formed by the memristor bridge proposed in [<a href="#B75-applsci-13-13309" class="html-bibr">75</a>]. The weight plot for Equation (6) is a straight line ranging from 0 to 1. The linear relationship between resistance and weight makes mathematical calculations easier, but requires a much larger number of memristors, which leads to additional resource costs.</p>
Full article ">Figure 5
<p>Architectures of the neuromorphic analog machine vision systems based on memristive devices. (<b>A</b>) For formal ANNs, weights are mapped to a crossbar array of memristive devices in cases where the bipolar weights are obtained due to a differential input. The input sensory circuits are connected directly to the inputs of neurons without digitalization. Visual information is encoded via the voltage amplitude, depending on illumination. (<b>B</b>) For spiking ANNs, presynaptic neurons generate spikes, the frequency of which depends on illumination. The value of the synaptic weights changes during the learning process. A postsynaptic neuron generates spikes when the charge on the membrane exceeds a threshold “Th”. So, the entire analog machine vision system is a spiking ANN without analog-to-digital and digital-to-analog converters.</p>
Full article ">Figure 6
<p>The SPICE model of a formal ANN circuit and an illustration of its operation. The initialization and programming signal generators, as well as the equivalent photodiode circuit, are on the left side. The memristor model and the OpAmp-based current-to-voltage convertor are on the right side. The current plot is shown for the TE point of the memristor. The plot shows two stages of the operation of the sensory part: the “write mode”—for recording the weight of an ANN synapse, and the “inference mode”—for scalar multiplication of the signal proportional to illumination according to the weight of an ANN synapse. Different colors correspond to different illumination levels or photocurrents—from 10 (pink) to 100 μA (green).</p>
Full article ">Figure 7
<p>The SPICE model of a spiking ANN circuit and the illustration of its operation. In comparison with the previous model (described in <a href="#applsci-13-13309-f006" class="html-fig">Figure 6</a>), an integrator with a threshold U<sub>4</sub> has been added, which converts illumination into pulse frequency. An additional amplifier U<sub>1</sub> and two additional switches S<sub>1</sub> and S<sub>7</sub> are used to amplify the pulses if the signal is used to potentiate or depress the weights of the first layer of a spiking ANN. The current plot shows that the frequency at which the pulses are fed into the memristor at the TE point is various—with a high frequency (pink color) for intensive illumination, and vice versa (green color).</p>
Full article ">Figure 8
<p>Experimental set-up for measuring data for simulation. (<b>A</b>) The chip consists of a linear 63 × 1 memristive array for the sensory part and a 32 × 8 memristive crossbar array for the ANN neurons. IV-curves of the memristive devices can be seen on the right. (<b>B</b>) Several parts are connected in the set-up: 1—PC; 2, 3, 6—PCB with the necessary electronic devices; 4—memristive chip; 5—light-proof container covering the light sources (LEDs) and the photodiodes (PDs).</p>
Full article ">Figure 9
<p>Computer modeling of a machine vision system with a formal ANN model. This figure shows examples of visual patterns that represent mathematical symbols. These images are fed to the input of a single-layer formal ANN. Each pixel of each image corresponds to a different signal amplitude. It can be seen that after training and transferring the synapse weights to the memristor resistances in the crossbar, each ANN neuron responds to the corresponding class of the visual pattern.</p>
Full article ">Figure 10
<p>Computer modeling of a computer vision system with a spiking model of an ANN. The architecture of this ANN is the same as for the formal ANN; however, this design contains integrators with a threshold and feedback from outputs to inputs. Feedbacks are active only during ANN training. During the training process of the ANN, the weights to which the high-frequency signal is applied tend to increase, and vice versa. Thus, ANN neurons learn to recognize visual patterns.</p>
Full article ">Figure 11
<p>Roadmap for further development in this area (DVS—dynamic vision sensor; IR—infrared).</p>
Full article ">
Back to TopTop