Search

The rapid increase of wearable sensor devices poses new challenges for implementing continuous real-time processing of physiological data. Neuromorphic sensory-processing devices can enable both the measurement of bio-signals and their processing locally in compact embedded wearable systems. In particular, mixed-signal spiking neural networks implemented on neuromorphic processors can be integrated directly with the sensors to extract temporal data-streams in real-time with very low power consumption. In this work, we present a neuromorphic approach for classifying spatio-temporal data from electromyography (EMG) signals, which paves the way toward the realization of compact wearable solutions for neuroprosthetic control. Here we extend previously proposed delta-encoding methods to transform bio-signals into spike trains and use a spiking Recurrent Neural Network (SRNN) architecture to extract features from them. The SRNN was first simulated in software to find the optimal set of hyperparameters, and then validated on the neuromorphic hardware, with a difference in the performance of less than 2%. We describe how biologically plausible mechanisms such as Spike-Timing Dependent Plasticity (STDP) and soft Winner-Take-All (WTA) networks can be exploited to classify the EMG signals and show how their combined use in EMG data classification achieves competitive results with different datasets. Specifically, the classification performance for the Roshambo EMG dataset, which has three different classes, is above 85%, and for the basic finger movements dataset from the Ninapro database, which has eight different classes, reaches 55% accuracy.

Wearable devices are a fast-growing technology with impact on personal healthcare for both society and economy. Due to the widespread of sensors in pervasive and distributed networks, power consumption, processing speed, and system adaptation are vital in future smart wearable devices. The visioning and forecasting of how to bring computation to the edge in smart sensors have already begun, with an aspiration to provide adaptive extreme edge computing. Here, we provide a holistic view of hardware and theoretical solutions toward smart wearable devices that can provide guidance to research in this pervasive computing era. We propose various solutions for biologically plausible models for continual learning in neuromorphic computing technologies for wearable sensors. To envision this concept, we provide a systematic outline in which prospective low power and low latency scenarios of wearable sensors in neuromorphic platforms are expected. We successively describe vital potential landscapes of neuromorphic processors exploiting complementary metal-oxide semiconductors (CMOS) and emerging memory technologies (e.g., memristive devices). Furthermore, we evaluate the requirements for edge computing within wearable devices in terms of footprint, power consumption, latency, and data size. We additionally investigate the challenges beyond neuromorphic computing hardware, algorithms and devices that could impede enhancement of adaptive edge computing in smart wearable devices., Frontiers in Neuroscience, 15, ISSN:1662-453X, ISSN:1662-4548

DNA and RNA polymerases (Pols) are central to life, health, and biotechnology because they allow the flow of genetic information in biological systems. Importantly, Pol function and (de)regulation are linked to human diseases, notably cancer (DNA Pols) and viral infections (RNA Pols) such as COVID-19. In addition, Pols are used in various applications such as synthesis of artificial genetic polymers and DNA amplification in molecular biology, medicine, and forensic analysis. Because of all of this, the field of Pols is an intense research area, in which computational studies contribute to elucidating experimentally inaccessible atomistic details of Pol function. In detail, Pols catalyze the replication, transcription, and repair of nucleic acids through the addition, via a nucleotidyl transfer reaction, of a nucleotide to the 3′-end of the growing nucleic acid strand. Here, we analyze how computational methods, including force-field-based molecular dynamics, quantum mechanics/molecular mechanics, and free energy simulations, have advanced our understanding of Pols. We examine the complex interaction of chemical and physical events during Pol catalysis, like metal-aided enzymatic reactions for nucleotide addition and large conformational rearrangements for substrate selection and binding. We also discuss the role of computational approaches in understanding the origin of Pol fidelity—the ability of Pols to incorporate the correct nucleotide that forms a Watson–Crick base pair with the base of the template nucleic acid strand. Finally, we explore how computations can accelerate the discovery of Pol-targeting drugs and engineering of artificial Pols for synthetic and biotechnological applications. This article is categorized under: Structure and Mechanism > Reaction Mechanisms and Catalysis Structure and Mechanism > Computational Biochemistry and Biophysics Software > Molecular Modeling. © 2021 The Authors. WIREs Computational Molecular Science published by Wiley Periodicals LLC.

An accurate description of muscular activity plays an important role in the clinical diagnosis and rehabilitation research. The electromyography (EMG) is the most used technique to make accurate descriptions of muscular activity. The EMG is associated with the electrical changes generated by the activity of the motor neurons. Typically, to decode the muscular activation during different movements, a large number of individual motor neurons are monitored simultaneously, producing large amounts of data to be transferred and processed by the computing devices. In this paper, we follow an alternative approach that can be deployed locally on the sensor side. We propose a neuromorphic implementation of a spiking neural network (SNN) to extract spatio-temporal information of EMG signals locally and classify hand gestures with very low power consumption. We present experimental results on the input data stream using a mixed-signal analog/digital neuromorphic processor. We performed a thorough investigation on the performance of the SNN implemented on the chip, by: first, calculating PCA on the activity of the silicon neurons at the input and the hidden layers to show how the network helps in separating the samples of different classes; second, performing classification of the data using state-of-the-art SVM and logistic regression methods and a hardware-friendly spike-based read-out. The traditional algorithm achieved a classification rate of [Formula: see text] and [Formula: see text], respectively, and the spiking learning method achieved [Formula: see text]. The power consumption of the SNN is [Formula: see text], showing the potential of this approach for ultra-low power processing.

Wearable sensor devices are disrupting healthcare technologies with a rapid increase of systems able to perform continuous monitoring of physiological data. These systems however typically produce large amounts of data that needs to be processed in real-time. Neuromorphic processors represent a promising class of devices that can be directly interfaced with the sensors to extract temporal data-streams in real-time with very low-power consumption. In this paper we present an application of such a neuromorphic approach for classifying electromyography (EMG) signals, that can be useful for designing compact wearable solutions for neuroprosthetic control. As the neuromorphic processor comprises configurable spiking neurons and adaptive synapses, we propose to use a spiking recurrent neural network (SRNN) for classifying the spatio-temporal data derived from the EMG signals. We performed a thorough investigation on the performance of the network implemented on the chip, by evaluating the classification performance of two hardware-friendly spike-based read-out models: a rate-based state distance model, and a spiking-time-dependent plasticity (STDP) learning model. The results were then compared with a classical machine learning approach based on the Support Vector Machine (SVM) method. Our results show how the classification accuracy of the state distance method is above 75%, which is even better than the SVM results; while the STDP learning rule only achieved 60% accuracy.

The advent of dedicated Deep Learning (DL) accelerators and neuromorphic processors has brought on new opportunities for applying both Deep and Spiking Neural Network (SNN) algorithms to healthcare and biomedical applications at the edge. This can facilitate the advancement of medical Internet of Things (IoT) systems and Point of Care (PoC) devices. In this paper, we provide a tutorial describing how various technologies including emerging memristive devices, Field Programmable Gate Arrays (FPGAs), and Complementary Metal Oxide Semiconductor (CMOS) can be used to develop efficient DL accelerators to solve a wide variety of diagnostic, pattern recognition, and signal processing problems in healthcare. Furthermore, we explore how spiking neuromorphic processors can complement their DL counterparts for processing biomedical signals. The tutorial is augmented with case studies of the vast literature on neural network and neuromorphic hardware as applied to the healthcare domain. We benchmark various hardware platforms by performing a sensor fusion signal processing task combining electromyography (EMG) signals with computer vision. Comparisons are made between dedicated neuromorphic processors and embedded AI accelerators in terms of inference latency and energy. Finally, we provide our analysis of the field and share a perspective on the advantages, disadvantages, challenges, and opportunities that various accelerators and neuromorphic processors introduce to healthcare and biomedical domains., IEEE Transactions on Biomedical Circuits and Systems, 14 (6), ISSN:1932-4545, ISSN:1940-9990

Living organisms are far superior to state-of-the-art robots as they have evolved a wide number of capabilities that far encompass our most advanced technologies. The merging of biological and artificial world, both physically and cognitively, represents a new trend in robotics that provides promising prospects to revolutionize the paradigms of conventional bio-inspired design as well as biological research. In this review, a comprehensive definition of animal-robot interactive technologies is given. They can be at animal level, by augmenting physical or mental capabilities through an integrated technology, or at group level, in which real animals interact with robotic conspecifics. Furthermore, an overview of the current state of the art and the recent trends in this novel context is provided. Bio-hybrid organisms represent a promising research area allowing us to understand how a biological apparatus (e.g. muscular and/or neural) works, thanks to the interaction with the integrated technologies. Furthermore, by using artificial agents, it is possible to shed light on social behaviours characterizing mixed societies. The robots can be used to manipulate groups of living organisms to understand self-organization and the evolution of cooperative behaviour and communication.

The demonstration shows a mobile application, called "RELAX", for hand gesture classification using multi-sensors fusion. In particular, we integrated the data collected by an electromyography (EMG) sensor with the events produced by an event-based vision sensor, the Dynamic Vision Sensor (DVS). The application runs real-time on any Android smartphone and it is able to recognize five gestures with an accuracy of up to 85%.This demonstration is associated with the track Bio-Inspired and Neuromorphic Circuits and Systems. Associated paper submission identifier: 8114.

Best demo Award; International audience; The discrimination of human gestures using wearable solutions is extremely important as a supporting technique for assisted living, healthcare of the elderly and neurorehabilitation. This paper presents a mobile electromyography (EMG)analysis framework to be an auxiliary component in physiotherapy sessions or as a feedback for neuroprosthesis calibration. We implemented a framework that allows the integration of multisensors, EMG and visual information, to perform sensor fusion and to improve the accuracy of hand gesture recognition tasks. In particular, we used an event-based camera adapted to run on the limited computational resources of mobile phones. We introduced a new publicly available dataset of sensor fusion for hand gesture recognition recorded from 10 subjects and used it to train the recognition models offline. We compare the online results of the hand gesture recognition using the fusion approach with the individual sensors with an improvement in the accuracy of 13% and 11%, for EMG and vision respectively, reaching 85%.

Electromyography (EMG) signals carry information about the movements of skeleton muscles. EMG on-line processing and analysis can be applied to different types of human-machine interfaces and provide advantages to patient rehabilitation strategies in case of injuries or stroke. However, continuous monitoring and data collection produces large amounts of data and introduces a bottleneck for further processing by computing devices. Neuromorphic technology offers the possibility to process the data directly on the sensor side in real-time, and with very low power consumption. In this work we present the first steps toward the design of a neuromorphic event-based neural processing system that can be directly interfaced to surface EMG (sEMG) sensors for the on-line classification of the motor neuron output activities. We recorded the EMG signals related to two movements of open and closed hand gestures, converted them into asynchronous Address-Event Representation (AER) signals, provided them in input to a recurrent spiking neural network implemented on an ultra-low power neuromorphic chip, and analyzed the chip's response. We configured the recurrent network as a Liquid State Machine (LSM) as a means to classify the spatio-temporal data and evaluated the Separation Property (SP) of the liquid states for the two movements. We present experimental results which show how the activity of the silicon neurons can be encoded in state variables for which the average state distance is larger between two different gestures than it is between the same ones measured across different trials.

An ever increasing amount of robotic platforms are being equipped with a new generation of neuromorphic computing architectures. Neuromorphic computing systems represent a promising brain-inspired technology that use asynchronous pulses to encode, transmit, and process sensory signals, typically implemented in compact low-latency and low-power devices. However, although multiple examples of sensing and processing neuromorphic devices have been successfully deployed on robotic platforms, no example of event-based neuromorphic motor controller has been proposed yet. In this paper, we present an open-loop neuromorphic controller implemented using a full-custom spiking neural network VLSI chip interfaced to motors for performing position control. The proposed controller is based on biologically inspired principles by which the discharge of motor-neuron populations produces muscle contractions. Following these principles, we use the spikes of the silicon neurons present in the neuromorphic chip to encode the target position and drive the motors using Pulse Frequency Modulation (PFM) technique, rather than the more traditional Pulse Width Modulation (PWM) one.

The use of robotics to establish social interactions between animals and robots, represents an elegant and innovative method to investigate animal behaviour. However, robots are still underused to investigate high complex and flexible behaviours, such as aggression. Here, Betta splendens was tested as model system to shed light on the effect of a robotic fish eliciting aggression. We evaluated how multiple signal systems, including a light stimulus, affect aggressive responses in B. splendens. Furthermore, we conducted experiments to estimate if aggressive responses were triggered by the biomimetic shape of fish replica, or whether any intruder object was effective as well. Male fishes showed longer and higher aggressive displays as puzzled stimuli from the fish replica increased. When the fish replica emitted its full sequence of cues, the intensity of aggression exceeded even that produced by real fish opponents. Fish replica shape was necessary for conspecific opponent perception, evoking significant aggressive responses. Overall, this study highlights that the efficacy of an artificial opponent eliciting aggressive behaviour in fish can be boosted by exposure to multiple signals. Optimizing the cue combination delivered by the robotic fish replica may be helpful to predict escalating levels of aggression.

The understanding of the phenomena at the base of tooth movement, due to orthodontic therapy, is an ambitious topic especially with regard to the "optimal forces" able to move teeth without causing irreversible tissue damages. To this aim, a measuring platform for detecting 3D orthodontic actions has been developed. It consists of customized load cells and dedicated acquisition electronics. The force sensors are able to detect, simultaneously and independently of each other, the six orthodontic components which a tooth is affected by. They have been calibrated and then applied on a clinical case that required NiTi closed coil springs and miniscrews for the treatment of upper post-extraction spaces closure. The tests have been conducted on teeth stumps belonging to a plaster cast of the patient's mouth. The load cells characteristics (sensor linearity and repeatability) have been analyzed (0.97 < R 2 < 1; 6.3*10 -6 % < STD < 8.8 %) and, on the basis of calibration data, the actions exerted on teeth have been determined. The biomechanical behavior of the frontal group and clinical interpretation of the results are discussed.

A robotic fish has been developed to create a mixed bio-hybrid system made up of weakly electric fish and a mobile dummy fish. Weakly electric fish are capable of interacting with each other via sequences of self-generated electric signals during electrocommunication. Here we present the design of an artificial dummy fish, which is subsequently tested in behavioural experiments. The robot consists of two parts: a flexible tail that can move at different frequencies and amplitudes, performing a carangiform oscillation, and a rigid head containing the motor for the tail oscillation. The dummy fish mimics the weakly electric fish Mormyrus rume in morphology, size and electric signal generation. In order to study electrical interactions, the dummy fish is equipped with ten electrodes that record electric signals of nearby real fish and generate electric dipole fields around itself that are similar to those produced by real fish in both waveform and sequence. Behavioural experiments demonstrate that the dummy fish is able to recruit both single individuals and groups of M. rume from a shelter into an exposed area. The development of an artificial dummy fish may help to understand fundamental aspects of collective behaviour in weakly electric fish and the properties necessary to initiate and sustain it in closed-loop feedback experiments based on electrocommunication.

The study proposed describes preliminary results of a spiking implementation of lamprey's Central Pattern Generator (CPG) using Neuromorphic VLSI devices. Several robotic lamprey implementations have been built to test the models in a bio-mimetic artifact but, in these systems there is a clear separation between the mechanical system, and their control part. This study aims to implement a CPG hardware network, to directly control actuators, creating a biomimetic robot both from mechanical and electronic point of view.

Lateralization (i.e. left-right asymmetries in the brain and behaviour) of courtship displays has been examined in a growing number vertebrate species, while evidence for invertebrates is limited. In this study, we investigated lateralization of courtship and mating displays in the parasitic wasp Leptomastidea abnormis. Results showed a population-level lateralization of male courtship displays. Male antennal tapping on the female's head was right-biased. However, right-biased male courtship acts were not characterized by higher male antennal tapping frequencies, nor success in mating although antennal tapping frequency was higher in males with mating success with respect to unsuccessful males. Overall, our results add basic knowledge to the behavioural ecology of insect parasitoids. To the best of our knowledge, this is the first report of behavioural lateralization in parasitic Hymenoptera.

Many hymenopterous parasitoids are biological control agents that contribute significant mortality to economically important arthropod pests of agriculture, forestry, and natural ecosystems. These critically important species forage for food, mates and hosts in complex environments, relying on a variety of stimuli including visual, olfactory, tactile and vibrational cues. In these contexts, learning is often key to optimizing fitness. Here we review current knowledge on the associative learning abilities of both immature and adult parasitoids, specifically learning associated with different kinds of stimuli: (i) reward-associated visual cues, such as colors, shapes and damage patterns; (ii) reward-associated chemical cues, including long- and short-range olfactory cues originating from the plant-host complex or from the host itself, as well as exogenous, nonhost-related odors, and (iii) danger-associated chemical cues that can trigger avoidance behavior. We highlight how the inclusion of associative training with visual and olfactory stimuli in mass-rearing protocols might be exploited to enhance the efficiency of biological control agents against a given target pest. We also discuss the genetic, physiological and ecological factors that can impact parasitoid learning and suggest ways they might be manipulated to optimize rearing conditions for biological control objectives.

The lamprey has been often used as a model for understanding the role of Central Pattern Generators (CPGs) in locomotion. Many artificial neural network models have been proposed in the past to explain the neuro-physiology and behavioral data measured from the lamprey, and several robotic implementations have been built to test the software models in a real physical bio-mimetic artifact, and to reproduce the characteristic locomotion patterns observed in the real lamprey. However, in these systems there has typically been a clear separation between the mechanical component of the system (the body), and its control part (the CPG), typically implemented with conventional digital platforms, such as micro-controllers or Field Programmable Gate Arrays (FPGAs). Here we propose to implement a CPG network using neuromorphic electronic circuits, that can be directly interfaced to the robotic actuators of a bio-mimetic robotic lamprey, eliminating the distinction between software and hardware. These circuits comprise low- power analog silicon neurons and synapses, that are affected by device mismatch and noise. The challenge is therefore to determine the CPG model that can best implement robust locomotion control of the bio-mimetic artifact, in face of the constraints imposed by the neuromorphic implementation. As these constraints are similar to the ones faced by the neurons and synapses in the real lamprey (e.g., finite and small power consumption, finite and small resolution or signal to noise ratio, large variability, etc.), the final system implementation will shed light onto the neural processing principles used by real CPG networks to produce robust and distributed control of locomotion in a physical bio-mimetic artifact.

A customized load cell is developed to quantitatively evaluate the three-dimensional orthodontic actions applied during treatment. The force-torque sensor is part of a platform composed of 14 load cells, each one equipped with 6 strain gauges and interfaced with a tooth. The particular shape of the load cell allows detecting 6 mechanical actions independently that the tooth is subjected to. At the same time, a dedicated acquisition system is used to collect data simultaneously. The load cell is calibrated by applying known loads in a range between 0 N and 2 N. For each strain gauge's output, good linearity (0.83 < R2 < 0.99) and great repeatability (10−4% < STD < 1.4%) are observed. Within this work, the load cell design, fabrication and characterization are described.

Introduction Orthodontic treatment is an important part of dental health care in Europe: the percentages of the population undergoing therapy vary from 10% to 55%. Therefore, quantifying effective orthodontic loads is a challenging topic with regard to the predictability of tooth movements and the reduction of traumatic side effects. Methods A customized measuring platform was developed and used for detecting orthodontic forces in a range between 0.1 and 2 N. The system consists of 6 load cells, each equipped with 6 strain gauges. The tests were conducted on a 3-dimensional printed malocclused mouth model and on a plaster cast. Four types of superelastic ligation and 2 types of invisible aligners were tested to analyze, respectively, a malocclusion with a high maxillary canine, and the effects on the axial rotation of a maxillary central incisor with and without a divot in the invisible aligners. Results Optimal treatment forces are exerted by low-friction wires, especially if they are partially engaged. Moreover, by reducing the treatment force, there is less necessity of anchoring to surrounding teeth, thus decreasing the side effects. The efficacy of using invisible aligners with a divot was validated. Conclusions This platform allowed measurement, at the radicular level, of the resultant forces of orthodontic treatments performed with different orthodontic appliances. In addition to customizing and calibrating the therapy for each patient, this platform could be used to develop new specific instruments able to exert lower treatment forces, thus preventing irreversible damages.

Objective. Aim of this study was to determine the prevalence and distribution of the Norovirus (NoVs) infections, from October 2007 to April 2008, in children hospitalized for gastroenteritis. Materials and methods. From October 2007 to April 2008 faecal specimens were obtained from 220 infants and children (130 males, 90 males) of whom 58 under 1, 60 between 1 and 2, 28 between 2 and 3, 74 between 3 and 12 years old, admitted for gastroenteritis to the Department of Paediatrics at the Campo di Marte’s Hospital in Lucca, Italy. All the specimens were examined both for the presence of bacteria (Salmonella spp, Shigella spp, Campylobacter spp, Yersinia spp, Aeromonas spp and enteropathogenic E. coli) and for the presence of rotaviruses (HRVs) and adenoviruses (AdVs). Sixty-eight frozen stool samples negative for bacteria and viruses were examined for NoVs. HRVs and AdVs were detected by the immunochromatographic test RIDAQUICK Rotavirus/Adenovirus Combi test (R-Biopharm, Germany). NoVs detection was carried out by the enzyme immunoassay RIDASCREEN Norovirus 3rd Generation EIA (R-Biopharm, Germany). In the RIDASCREEN Norovirus test, specific monoclonal antibodies against antigens of several different genotypes are used in a sandwich type method. In a validation study of the RIDASCREEN Norovirus ELISA 3rd Generation at the Institute of Virology of the University of Dresden, the test correlation with the RT PCR showed the following performance, sensitivity (83.0%), specificity (100.0%), PPV (100.0%), NPV (83.0%). Results. HRVs were detected in 41.8% (89/220), AdVs in 4.1% (9/220) and Enteropathogenic bacteria were detected only in 7 (3.2%) patient. NoVs were detected in 18 of 68 studied cases (26.5%). Eighty-three of the NoVs and 49% of the HRVs infections occurred in children up two years old. Conclusions. The present results point out the importance of NoVs infections in childhood hospitalization, mainly in the first two years of life.The study shows the need to implement norovirus ELISA detection assays for clinical diagnosis and for planning an active surveillance among paediatric cases requiring hospitalization due to gastroenteritis for the development of effective preventive measures.