Search

Microwave imaging has been investigated for medical applications such as stroke and breast imaging. Current systems typically rely on bench-top equipment to scan at a variety of antenna positions. For dynamic imaging of moving structures, such as the cardiovascular system, much higher imaging speeds are required than what has thus far been reported. Recent innovations in radar-on-chip technology allow for simultaneous high speed data collection at multiple antenna positions at a fraction of the cost of conventional microwave equipment, in a small and potentially portable system. The objective of the current work is to provide proof of concept of dynamic microwave imaging in the body, using radar-on-chip technology.Arrays of body-coupled antennas were used with nine simultaneously operated coherent ultra-wideband radar chips. Data were collected from the chest and thigh of a volunteer, with the objective of imaging the femoral artery and beating heart. In addition, data were collected from a phantom to validate system performance. Video data were constructed using beamforming.The location of the femoral artery could successfully be resolved, and a distinct arterial pulse wave was discernable. Cardiac activity was imaged at locations corresponding to the heart, but image quality was insufficient to identify individual anatomical structures. Static and differential imaging of the femur bone proved unsuccessful.Using radar chip technology and an imaging approach, cardiovascular activity was detected in the body, demonstrating first steps towards biomedical dynamic microwave imaging. The current portable and modular system design was found unsuitable for static in-body imaging.This first proof of concept demonstrates that radar-on-chip could enable cardiovascular imaging in a low-cost, small and portable system. Such a system could make medical imaging more accessible, particularly in ambulatory or long-term monitoring settings.

Recent translational efforts in brain-machine interfaces (BMI) are demonstrating the potential to help people with neurological disorders. The current trend in BMI technology is to increase the number of recording channels to the thousands, resulting in the generation of vast amounts of raw data. This in turn places high bandwidth requirements for data transmission, which increases power consumption and thermal dissipation of implanted systems. On-implant compression and/or feature extraction are therefore becoming essential to limiting this increase in bandwidth, but add further power constraints – the power required for data reduction must remain less than the power saved through bandwidth reduction. Spike detection is a common feature extraction technique used for intracortical BMIs. In this paper, we develop a novel firing-rate-based spike detection algorithm that requires no external training and is hardware efficient and therefore ideally suited for real-time applications. Key performance and implementation metrics such as detection accuracy, adaptability in chronic deployment, power consumption, area utilization, and channel scalability are benchmarked against existing methods using various datasets. The algorithm is first validated using a reconfigurable hardware (FPGA) platform and then ported to a digital ASIC implementation in both 65 nm and 0.18MU m CMOS technologies. The 128-channel ASIC design implemented in a 65 nm CMOS technology occupies 0.096 mm2 silicon area and consumes 4.86MU W from a 1.2 V power supply. The adaptive algorithm achieves a 96% spike detection accuracy on a commonly used synthetic dataset, without the need for any prior training.

Extracellular recordings are typically analysed by separating them into two distinct signals: local field potentials (LFPs) and spikes. Previous studies have shown that spikes, in the form of single-unit activity (SUA) or multiunit activity (MUA), can be inferred solely from LFPs with moderately good accuracy. SUA and MUA are typically extracted via threshold-based technique which may not be reliable when the recordings exhibit a low signal-to-noise ratio (SNR). Another type of spiking activity, referred to as entire spiking activity (ESA), can be extracted by a threshold-less, fast, and automated technique and has led to better performance in several tasks. However, its relationship with the LFPs has not been investigated. In this study, we aim to address this issue by inferring ESA from LFPs intracortically recorded from the motor cortex area of three monkeys performing different tasks. Results from long-term recording sessions and across subjects revealed that ESA can be inferred from LFPs with good accuracy. On average, the inference performance of ESA was consistently and significantly higher than those of SUA and MUA. In addition, local motor potential (LMP) was found to be the most predictive feature. The overall results indicate that LFPs contain substantial information about spiking activity, particularly ESA. This could be useful for understanding LFP-spike relationship and for the development of LFP-based BMIs.

Objective. Translational efforts on spike-signal-based implantable brain-machine interfaces (BMIs) are increasingly aiming to minimise bandwidth while maintaining decoding performance. Developing these BMIs requires advances in neuroscience and electronic technology, as well as using low-complexity spike detection algorithms and high-performance machine learning models. While some state-of-the-art BMI systems jointly design spike detection algorithms and machine learning models, it remains unclear how the detection performance affects decoding. Approach. We propose the co-design of the neural decoder with an ultra-low complexity spike detection algorithm. The detection algorithm is designed to attain a target firing rate, which the decoder uses to modulate the input features preserving statistical invariance in long term (over several months). Main results. We demonstrate a multiplication-free fixed-point spike detection algorithm with an average detection accuracy of 97% across different noise levels on a synthetic dataset and the lowest hardware complexity among studies we have seen. By co-designing the system to incorporate statistically invariant features, we observe significantly improved long-term stability, with decoding accuracy degrading by less than 10% after 80 days of operation. Our analysis also reveals a nonlinear relationship between spike detection and decoding performance. Increasing the detection sensitivity improves decoding accuracy and long-term stability, which means the activity of more neurons is beneficial despite the detection of more noise. Reducing the spike detection sensitivity still provides acceptable decoding accuracy whilst reducing the bandwidth by at least 30%. Significance. Our findings regarding the relationship between spike detection and decoding performance can provide guidance on setting the threshold for spike detection rather than relying on training or trial-and-error. The trade-off between data bandwidth and decoding performance can be effectively managed using appropriate spike detection settings. We demonstrate improved decoding performance by maintaining statistical invariance of input features. We believe this approach can motivate further research focused on improving decoding performance through the manipulation of data itself (based on a hypothesis) rather than using more complex decoding models.

Mass characterisation of emerging memory devices is an essential step in modelling their behaviour for integration within a standard design flow for existing integrated circuit designers. This work develops a novel characterisation platform for emerging resistive devices with a capacity of up to 1 million devices on-chip. Split into four independent sub-arrays, it contains on-chip column-parallel DACs for fast voltage programming of the DUT. On-chip readout circuits with ADCs are also available for fast read operations covering 5-decades of input current (20nA to 2mA). This allows a device's resistance range to be between 1k$\Omega$ and 10M$\Omega$ with a minimum voltage range of $\pm$1.5V on the device., Comment: 5 pages. To be published in ISCAS 2022 and made available on IEEE Xplore

Simulations of electroneurogram recording could help find the optimal set of electrodes and algorithms for selective neural recording. However, no flexible methods are established for selective neural recording as for neural stimulation. This paper proposes a method to couple a compartmental and a FEM nerve model, implemented in NEURON and COMSOL, respectively, to translate Node of Ranvier currents into extraneural electric fields. The study simulate ex-vivo experimental conditions, and the method allows flexibility in electrode geometries and nerve topologies. This model has been made available in a public repository

\normalsize The capability of in-memory computation, reconfigurability, low power operation as well as multistate operation of the memristive device deems them a suitable candidate for designing electronic circuits with a broad range of applications. Besides, the integrability of memristor with CMOS enables it to use in logic circuits too. In this work, we demonstrate with examples the design flow for memristor-based electronics, after the custom memristor model already being integrated and validated into our chosen Computer-Aided Design (CAD) tool to performing layout-versus-schematic and post-layout checks including the memristive device. We envisage that this step-by-step guide to introducing memristor into the standard integrated circuit design flow will be a useful reference document for both device developers who wish to benchmark their technologies and circuit designers who wish to experiment with memristive-enhanced systems.

Memristive technology has experienced explosive growth in the last decade, with multiple device structures being developed for a wide range of applications. However, transitioning the technology from the lab into the marketplace requires the development of an accessible and user-friendly design flow, supported by an industry-grade tool chain. In this work, we demonstrate the behaviour of our in-house fabricated custom memristor model and its integration into the Cadence Electronic Design Automation (EDA) tools for verification. Various input stimuli were given to record the memristive device characteristics both at the device level as well as the schematic level for verification of the memristor model. This design flow from device to industrial level EDA tools is the first step before the model can be used and integrated with Complementary Metal-Oxide Semiconductor (CMOS) in applications for hybrid memristor/CMOS system design.

The total economic cost of neurological disorders exceeds £100 billion per annum in the United Kingdom alone, yet pharmaceutical companies continue to cut investments due to failed clinical studies and risk [1]. These challenges motivate an alternative to solely pharmacological treatments. The emerging field of bioelectronics suggests a novel alternative to pharmaceutical intervention that uses electronic hardware to directly stimulate the nervous system with physiologically inspired electrical signals [2]. Given the processing capability of electronics and precise targeting of electrodes, the potential advantages of bioelectronics include specificity in the time, method, and location of treatment, with the ability to iteratively refine and update therapy algorithms in software [3]. A primary disadvantage of the current systems is invasiveness due to surgical implantation of the device.

The introduction of large-scale chemical sensing systems in CMOS which integrate millions of ISFET sensors have allowed applications such as DNA sequencing and fine-pixel chemical imaging systems to be realised. Using CMOS ISFETs provides advantages of digitisation directly at the sensor as well as correcting for non-linearity in its response. However, for this to be beneficial and scale, the readout circuits need to have the minimum possible footprint and power consumption. Within this context, this paper analyses an ISFET based pH-to-time readout using an inverter in the time-domain as a level-crossing detector and presents a 32 × 32 array with in-pixel digitisation for pH sensing. The inverter-based sensing pixel, controlled by a triangular waveform, converts the pH response into a time-domain signal whilst also compensating for sensor offset and thus resulting in an increase in dynamic range. The sensor pixels interface to a 15-bit asynchronous column-wise time-to-digital converter (TDC), enabling fast asynchronous conversion whilst using minimal silicon area. Parallel outputs of 32 TDC interfaces are serialised to achieve fast data throughput. This system is implemented in a standard 0.18 $\,\mu$ m CMOS technology, with a pixel size of 26 $\mu$ m × 26 $\mu$ m and a TDC area of 26 $\mu$ m × 180 $\mu$ m. Additionally, we investigate the use of additional offset compensation by having half of the array implemented with the floating gate tied down via a well diode. Measured results demonstrate the system is able to sense reliably with an average pH sensitivity of 30 mV/pH, whilst being able to compensate for sensor offset by up to $\pm$ 7 V. A resolution of 0.013 pH is achieved and noise measurements show an integrated noise of 0.08 pH within 2–500 Hz and SFDR of 42.6 dB. The total power consumption of the system is measured to be 11.286 mW when operating at a high frame rate of 1 KFPS.

Implantable brain machine interfaces (BMIs) are now on a trajectory to go mainstream, wherein what was once considered last resort will progressively become elective at earlier stages in disease treatment. First-in-human successes have demonstrated the ability to decode highly dexterous motor skills such as handwriting, and speech from human cortical activity. These have been used for cursor and prosthesis control, direct-to-text communication and speech synthesis. Along with these breakthrough studies, technology advancements have enabled the observation of more channels of neural activity through new concepts for centralised/distributed implant architectures. This is complemented by research in flexible substrates, packaging, surgical workflows and data processing. New regulatory guidance and funding has galvanised the field. This culmination of resource, efforts and capability is now attracting significant investment for BMI commercialisation. This paper reviews recent developments and describes the paradigm shift in BMI development that is leading to new innovations, insights and BMI translation.

Advancements in miniaturisation and new capabilities of implantable devices impose a need for the development of compact, hermetic, and CMOS-compatible micro packaging methods. Gold-tin-based eutectic bonding presents the potential for achieving low-footprint seals with low permeability to moisture at process temperatures below 350\degree C. In this paper, a method for the deposition of gold-tin eutectic alloy frames via sequential electroplating from commercially available solutions, with no special fabrication process, is described in detail. Bond quality was characterised through shear force measurements, scanning electron microscopy, visual inspection, and immersion tests. Characterisation of seals geometry, solder thickness, and bonding process parameters were evaluated, along with toxicity assessment of bonding layers to the human fibroblast cells. With a successful bond yield of over 70\% and no cytotoxic effect, AuSn eutectic bonding appears as a suitable method for the protection of integrated circuitry in implantable applications., 16 pages, 12 figures, paper submitted to Journal of Micromechanics and Microengineering

When mice are exposed to external warmth, nitric oxide synthase (NOS1) neurons in the median and medial preoptic (MnPO/MPO) hypothalamus induce sleep and concomitant body cooling. However, how these neurons regulate baseline sleep and body temperature is unknown. Using calcium photometry, we show that NOS1 neurons in MnPO/MPO are predominantly NREM active. This is the first instance of a predominantly NREM-active population in the PO area, or to our knowledge, elsewhere in the brain. In addition to releasing nitric oxide, NOS1 neurons in MnPO/MPO can release GABA, glutamate and peptides. We expressed tetanus-toxin light-chain in MnPO/MPO NOS1 cells to reduce vesicular release of transmitters. This induced changes in sleep structure: over 24 hours, mice had less NREM sleep in their dark (active) phase, and more NREM sleep in their light (sleep) phase. REM sleep episodes in the dark phase were longer, and there were fewer REM transitions between other vigilance states. REM sleep had less theta power. Mice with synaptically blocked MnPO/MPO NOS1 neurons were also warmer. In particular, mice were warmer than control mice at the dark-light transition (ZT0), as well as during the dark phase siesta (ZT16-20), where there is usually a body temperature dip. Also, at this siesta point of cooled body temperature, mice usually have more NREM, but mice with synaptically blocked MnPO/MPO NOS1 cells showed reduced NREM sleep at this time. Overall, MnPO/MPO NOS1 neurons promote both NREM and REM sleep and contribute to chronically lowering body temperature, particularly at transitions where the mice normally enter NREM sleep., bioRxiv

The unobtrusive monitoring of vital signals and behaviour can be used to gather intelligence to support the care of people living with dementia. This can provide insights into the person's wellbeing and the neurogenerative process, as well as enable them to continue to live safely at home, thereby improving their quality of life. Within this context, this study investigated the deployability of non-contact respiration rate (RR) measurement based on an Ultra-Wideband (UWB) radar System-on-Chip (SoC). An algorithm was developed to simultaneously and continuously extract the respiration signal, together with the confidence level of the respiration signal and the target position, without needing any prior calibration. The radar-measured RR results were compared to the RR results obtained from a ground truth measure based on the breathing sound, and the error rates were within 8% with a mean value of 2.5%. The target localisation results match to the radar-to-chest distances with a mean error rate of 5.8%. The tested measurement range was up to 5m. The results suggest that the algorithm could perform sufficiently well in non-contact stationary respiration rate detection.

The lack of open-source memory compilers in academia typically causes significant delays in research and design implementations. This paper presents an open-source memory compiler that is directly integrated within the Cadence Virtuoso environment using physical verification tools provided by Mentor Graphics (Calibre). It facilitates the entire memory generation process from netlist generation to layout implementation, and physical implementation verification. To the best of our knowledge, this is the first open-source memory compiler that has been developed specifically to automate Resistive Random Access Memory (RRAM) generation. RRAM holds the promise of achieving high speed, high density and non-volatility. A novel RRAM architecture, additionally is proposed, and a number of generated RRAM arrays are evaluated to identify their worst case control line parasitics and worst case settling time across the memristors of their cells. The total capacitance of lines SEL, N and P is 5.83 fF/cell, 3.31 fF/cell and 2.48 fF/cell respectively, while the total calculated resistance for SEL is 1.28 Ohm/cell and 0.14 Ohm/cell for both N and P lines., 4 pages

This paper describes Tiresias, a low-cost, unobtrusive networked radar system designed to monitor vulnerable patients in domestic environments and provide high quality behavioural and health data. Dementia is a disease that affects millions worldwide and progressively degrades an individual's ability to care for themselves. Eventually most people living with dementia will need to reside in assisted living facilities as they become unable to care for themselves. Understanding the effects dementia has on ability to self-care and extending the length of time people living with dementia can remain living independently are key goals of dementia research and care. The networked radar system proposed in this paper is designed to provide high quality behavioural and health data from domestic environments. This is achieved using multiple radar sensors networked together with their data outputs integrated and processed to produce high confidence measures of position and movement. It is hoped the data produced by this system will both provide insights into how dementia progresses, and also help monitor vulnerable individuals in their own homes, allowing them to remain independent longer than would otherwise be possible.

This paper investigates the relationship between Multi-Unit Activity (MUA) Binning Period (BP) and Brain-Computer Interface (BCI) decoding performance using Long-Short Term Memory decoders. The motivation is to determine whether lossy compression of MUA via increasing BP has any adverse consequences for BCI Behavioral Decoding Performance (BDP). The Neural data originates from intracortical recordings from Macaque Primary Motor cortex [1]. The BDP is measured by the Pearson correlation r between the observed and predicted velocity of the subject’s X-Y hand coordinates in reaching tasks [1]. The results suggest a statistically significant but slight linear relationship between increasing MUA BP and decreasing BDP. For example, when using a 100 ms moving average window, increasing the BP by 10 ms on average reduces the BDP r by approximately 0.85%. This relationship may be due to the reduced number of training examples, or due to the loss of Behavioral information because of reduced MUA temporal resolution.

This paper describes preliminary work towards an automated algorithm for labelling Neuropixel data that exploits the fact that adjacent recording sites are spatially oversampled. This is achieved by combining classical single channel spike sorting with spatial spike grouping, resulting in an improvement in both accuracy and robustness. This is additionally complemented by an automated method for channel selection that determines which channels contain high quality data. The algorithm has been applied to a freely accessible dataset, produced by Cortex Lab, UCL. This has been evaluated to have a accuracy of over 77% compared to a manually curated ground truth.

There is growing interest in using adaptive neuro-modulation to provide a more personalized therapy experience that might improve patient outcomes. Current implant technology, however, can be limited in its adaptive algorithm capability. To enable exploration of adaptive algorithms with chronic implants, we designed and validated the, ‘Picostim DyNeuMo Mk-1’, (DyNeuMo Mk-1 for short), a fully-implantable, adaptive research stimulator that titrates stimulation based on circadian rhythms (e.g. sleep, wake) and the patient’s movement state (e.g. posture, activity, shock, free-fall). The design leverages off-the-shelf consumer technology that provides inertial sensing with low-power, high reliability, and relatively modest cost. The DyNeuMo Mk-1 system was designed, manufactured and verified using ISO 13485 design controls, including ISO 14971 risk management techniques to ensure patient safety, while enabling novel algorithms. The system was validated for an intended use case in movement disorders under an emergency-device authorization from the Medicines and Healthcare Products Regulatory Agency (MHRA). The algorithm configurability and expanded stimulation parameter space allows for a number of applications to be explored in both central and peripheral applications. Intended applications include adaptive stimulation for movement disorders, synchronizing stimulation with circadian patterns, and reacting to transient inertial events such as posture changes, general activity, and walking. With appropriate design controls in place, first-in-human research trials are now being prepared to explore the utility of automated motion-adaptive algorithms.

Memory compilers are necessary tools to boost the design procedure of digital circuits. However, only a few are available to academia. Resistive Random Access Memory (RRAM) is characterised by high density, high speed, non volatility and is a potential candidate of future digital memories. To the best of the authors' knowledge, this paper presents the first open source RRAM compiler for automatic memory generation including its peripheral circuits, verification and timing characterisation. The RRAM compiler is written with Cadence SKILL programming language and is integrated in Cadence environment. The layout verification procedure takes place in Siemens Mentor Calibre tool. The technology used by the compiler is TSMC 180nm. This paper analyses the novel results of a plethora of M x N RRAMs generated by the compiler, up to M = 128, N = 64 and word size B = 16 bits, for clock frequency equal to 12.5 MHz. Finally, the compiler achieves density of up to 0.024 Mb/mm2., Comment: Final Version of NEWCAS 2022. 5 pages

The unobtrusive monitoring of vital signals and behaviour can be used to gather intelligence to support the care of people living with dementia. This can provide insights into the persons wellbeing and the neurogenerative process, as well as enable them to continue to live safely at home, thereby improving their quality of life. Within this context, this study investigated the deployability of non-contact respiration rate (RR) measurement based on an Ultra-Wideband (UWB) radar System-on-Chip (SoC). An algorithm was developed to simultaneously and continuously extract the respiration signal, together with the confidence level of the respiration signal and the target position, without needing any prior calibration. The radar-measured RR results were compared to the RR results obtained from a ground truth measure based on the breathing sound, and the error rates were within 8% with a mean value of 2.4%. The target localisation results match to the radarto-chest distances with a mean error rate of 5.4%. The tested measurement range was up to 5m. The results suggest that the algorithm could perform sufficiently well in non-contact stationary respiration rate detection.

Electrical neurostimulation is effective in treating neurological disorders, but associated recording artefacts generally limit applications to ‘open-loop’ stimuli. Since light does not prevent concurrent electrical recordings, optogenetics enables real-time, continuous ‘closed-loop’ control of brain activity. Here we show that closed-loop optogenetic stimulation with excitatory opsins (CLOSe) affords precise manipulation of neural dynamics, both in vitro, in brain slices from transgenic mice, and in vivo, with anesthetised monkeys. We demonstrate the generation of oscillations in quiescent tissue, enhancement or suppression of endogenous patterns in active tissue, and modulation of seizure-like bursts elicited by 4-aminopyridine. New network properties, emergent under CLOSe, depended on the phase-shift imposed between neural activity and optical stimulation, and could be modelled with a nonlinear dynamical system. In particular, CLOSe could stabilise or destabilise limit cycles associated with seizure oscillations, evident from systematic changes in the variability and entropy of seizure trajectories that correlated with their altered duration and intensity. Furthermore, CLOSe was achieved using intracortical optrodes incorporating light-emitting diodes, paving the way for translation of closed-loop optogenetics towards therapeutic applications in humans.

There is growing interest in using adaptive neuro-modulation to provide a more personalized therapy experience that might improve patient outcomes. Current implant technology, however, can be limited in its adaptive algorithm capability. To enable exploration of adaptive algorithms with chronic implants, we designed and validated the ‘Picostim DyNeuMo Mk-1’ (DyNeuMo Mk-1 for short), a fully-implantable, adaptive research stimulator that titrates stimulation based on circadian rhythms (e.g. sleep, wake) and the patient’s movement state (e.g. posture, activity, shock, free-fall). The design leverages off-the-shelf consumer technology that provides inertial sensing with low-power, high reliability, and relatively modest cost. The DyNeuMo Mk-1 system was designed, manufactured and verified using ISO 13485 design controls, including ISO 14971 risk management techniques to ensure patient safety, while enabling novel algorithms. The system was validated for an intended use case in movement disorders under an emergency-device authorization from the Medicines and Healthcare Products Regulatory Agency (MHRA). The algorithm configurability and expanded stimulation parameter space allows for a number of applications to be explored in both central and peripheral applications. Intended applications include adaptive stimulation for movement disorders, synchronizing stimulation with circadian patterns, and reacting to transient inertial events such as posture changes, general activity, and walking. With appropriate design controls in place, first-in-human research trials are now being prepared to explore the utility of automated motion-adaptive algorithms.

Chronic in-vivo neurophysiology experiments require highly miniaturized, remotely powered multi-channel neural interfaces which are currently lacking in power or flexibility post implantation. In this article, to resolve this problem we present the SenseBack system, a post-implantation reprogrammable wireless 32-channel bidirectional neural interfacing that can enable chronic peripheral electrophysiology experiments in freely behaving small animals. The large number of channels for a peripheral neural interface, coupled with fully implantable hardware and complete software flexibility enable complex in-vivo studies where the system can adapt to evolving study needs as they arise. In complementary ex-vivo and in-vivo preparations, we demonstrate that this system can record neural signals and perform high-voltage, bipolar stimulation on any channel. In addition, we demonstrate transcutaneous power delivery and Bluetooth 5 data communication with a PC. The SenseBack system is capable of stimulation on any channel with p20 V of compliance and up to 315 μA of current, and highly configurable recording with per-channel adjustable gain and filtering with 8 sets of 10-bit ADCs to sample data at 20 kHz for each channel. To the best of our knowledge this is the first such implantable research platform offering this level of performance and flexibility post-implantation (including complete reprogramming even after encapsulation) for small animal electrophysiology. Here we present initial acute trials, demonstrations and progress towards a system that we expect to enable a wide range of electrophysiology experiments in freely behaving animals.

Animals undertake specific behaviors before sleep. Little is known about whether these innate behaviors, such as nest building, are actually an intrinsic part of the sleep-inducing circuitry. We found, using activity-tagging genetics, that mouse prefrontal cortex (PFC) somatostatin/GABAergic (SOM/GABA) neurons, which become activated during sleep deprivation, induce nest building when opto-activated. These tagged neurons induce sustained global NREM sleep if their activation is prolonged metabotropically. Sleep-deprivation-tagged PFC SOM/GABA neurons have long-range projections to the lateral preoptic (LPO) and lateral hypothalamus (LH). Local activation of tagged PFC SOM/GABA terminals in LPO and the LH induced nesting and NREM sleep respectively. Our findings provide a circuit link for how the PFC responds to sleep deprivation by coordinating sleep preparatory behavior and subsequent sleep.

This paper investigates to what extent Long Short-Term Memory (LSTM) decoders can use Local Field Potentials (LFPs) to predict Single-Unit Activity (SUA) in Macaque Primary Motor cortex. The motivation is to determine to what degree the LFP signal can be used as a proxy for SUA, for both neuroscience and Brain-Computer Interface (BCI) applications. Firstly, the results suggest that the prediction quality varies significantly by implant location or animal. However, within each implant location / animal, the prediction quality seems to be correlated with the amount of power in certain LFP frequency bands (0-10, 10-20 and 40-50Hz, standardised LFPs). Secondly, the results suggest that bipolar LFPs are more informative as to SUA than unipolar LFPs. This suggests common mode rejection aids in the elimination of non-local neural information. Thirdly, the best individual bipolar LFPs generally perform better than when using all available unipolar LFPs. This suggests that LFP channel selection may be a simple but effective means of lossy data compression in Wireless Intracortical LFP-based BCIs. Overall, LFPs were moderately predictive of SUA, and improvements can likely be made.

Multiunit activity (MUA) has been proposed to mitigate the robustness issue faced by single-unit activity (SUA)-based brain-machine interfaces (BMIs). Most MUA-based BMIs still employ a binning method for extracting firing rates and linear decoder for decoding behavioural parameters. The limitations of binning and linear decoder lead to suboptimal performance of MUA-based BMIs. To address this issue, we propose Bayesian adaptive kernel smoother (BAKS) as the feature extraction method and long short-term memory (LSTM)-based deep learning as the decoding algorithm. We evaluated the proposed methods for reconstructing (offline) hand kinematics from intracortical neural data chronically recorded from the motor cortex of a monkey. Experimental results showed that BAKS coupled with LSTM outperformed other combinations of feature extraction method (binning or fixed kernel smoother) and decoding algorithm (Kalman filter or Wiener filter). Overall results demonstrate the effectiveness of BAKS and LSTM for improving the decoding performance of MUA-based BMIs.

ObjectiveThere has recently been an increasing interest in local field potential (LFP) for brain-machine interface (BMI) applications due to its desirable properties (signal stability and low bandwidth). LFP is typically recorded with respect to a single unipolar reference which is susceptible to common noise. Several referencing schemes have been proposed to eliminate the common noise, such as bipolar reference, current source density (CSD), and common average reference (CAR). However, to date, there have not been any studies to investigate the impact of these referencing schemes on decoding performance of LFP-based BMIs.ApproachTo address this issue, we comprehensively examined the impact of different referencing schemes and LFP features on the performance of hand kinematic decoding using a deep learning method. We used LFPs chronically recorded from the motor cortex area of a monkey while performing reaching tasks.Main resultsExperimental results revealed that local motor potential (LMP) emerged as the most informative feature regardless of the referencing schemes. Using LMP as the feature, CAR was found to yield consistently better decoding performance than other referencing schemes over long-term recording sessions.SignificanceOverall, our results suggest the potential use of LMP coupled with CAR for enhancing the decoding performance of LFP-based BMIs.

Extracellular recordings are typically analysed by separating them into two distinct signals: local field potentials (LFPs) and spikes. Understanding the relationship between these two signals is essential for gaining deeper insight into neuronal coding and information processing in the brain and is also relevant to brain-machine interface (BMI) research. Previous studies have shown that spikes, in the form of single-unit activity (SUA) or multiunit activity (MUA), can be inferred solely from LFPs with moderately good accuracy. These spiking activities that are typically extracted via threshold-based technique may not be reliable when the recordings exhibit a low signal-to-noise ratio (SNR). Another spiking activity in the form of a continuous signal, referred to as entire spiking activity (ESA), can be extracted by a threshold-less, fast, and automated technique and has led to better performance in several tasks. However, its relationship with the LFPs has not been investigated. In this study, we aim to address this issue by employing a deep learning method to infer ESA from LFPs intracortically recorded from the motor cortex area of two monkeys performing different tasks. Results from long-term recording sessions and across different tasks revealed that the inference accuracy of ESA yielded consistently and significantly higher accuracy than that of SUA and MUA. In addition, local motor potential (LMP) was found to be the most highly predictive feature compared to other LFP features. The overall results indicate that LFPs contain substantial information about the spikes, particularly ESA, which could be useful for the development of LFP-based BMIs. The results also suggest the potential use of ESA as an alternative neuronal population activity measure for analysing neural responses to stimuli or behavioural tasks.

This paper investigates the effectiveness of four Huffman-based compression schemes for different intracortical neural signals and sample resolutions. The motivation is to find effective lossless, low-complexity data compression schemes for Wireless Intracortical Brain-Machine Interfaces (WI-BMI). The considered schemes include pre-trained Lone 1st and 2nd order encoding [1], pre-trained Delta encoding, and pre-trained Linear Neural Network Time (LNNT) encoding [2]. Maximum codeword-length limited versions are also considered to protect against overfit to training data. The considered signals are the Extracellular Action Potential signal, the Entire Spiking Activity signal, and the Local Field Potential signal. Sample resolutions of 5 to 13 bits are considered. The result show that overfit-protection dramatically improves compression, especially at higher sample resolutions. Across signals, 2nd order encoding generally performed best at lower sample resolutions, and 1st order, Delta and LNNT encoding performed best at higher sample resolutions. The proposed methods should generalise to other remote sensing applications where the distribution of the sensed data can be estimated a priori.

Brain-machine Interfaces (BMI) hold great potential for treating neurological disorders such as epilepsy. Technological progress is allowing for a shift from open-loop, pacemaker-class, intervention towards fully closed-loop neural control systems. Low power programmable processing systems are therefore required which can operate within the thermal window of 2° C for medical implants and maintain long battery life. In this work, we have developed a low power neural engine with an optimized set of algorithms which can operate under a power cycling domain. We have integrated our system with a custom-designed brain implant chip and demonstrated the operational applicability to the closed-loop modulating neural activities in in-vitro and in-vivo brain tissues: the local field potentials can be modulated at required central frequency ranges. Also, both a freely-moving non-human primate (24-hour) and a rodent (1-hour) in-vivo experiments were performed to show system reliable recording performance. The overall system consumes only 2.93 mA during operation with a biological recording frequency 50 Hz sampling rate (the lifespan is approximately 56 hours). A library of algorithms has been implemented in terms of detection, suppression and optical intervention to allow for exploratory applications in different neurological disorders. Thermal experiments demonstrated that operation creates minimal heating as well as battery performance exceeding 24 hours on a freely moving rodent. Therefore, this technology shows great capabilities for both neuroscience in-vitro/in-vivo applications and medical implantable processing units.

Within the rapidly advancing field of active implantable medical devices, power management is a major consideration. Devices that provide life critical (or avoiding life threatening) function require a dependable, always-on power source, for example pacemakers. There is then a trade-off with battery lifetime as to whether such devices employ a primary cell or rechargeable battery. With new applications requiring multi-module implants, there is now also a need for transmitting within the body from one device to another. This paper outlines the key considerations and the process to define and optimise the power management strategy. We then apply this to a case study application – developing an implanted, multi-module closed-loop neuromodulation device for the treatment of focal epilepsy.

Next generation implantable neural interfaces are targeting devices with mm-scale form factors that are freely floating and completely wireless. Scalability to more recording (or stimulation) channels will be achieved through distributing multiple devices, instead of the current approach that uses a single centralized implant wired to individual electrodes or arrays. In this way, challenges associated with tethers, micromotion, and reliability of wiring is mitigated. This concept is now being applied to both central and peripheral nervous system interfaces. One key requirement, however, is to maximize specific absorption rate (SAR) constrained achievable wireless power transfer efficiency (PTE) of these inductive links with mm-sized receivers. Chip-scale coil structures for microsystem integration that can provide efficient near-field coupling are investigated. We develop near-optimal geometries for three specific coil structures: in-CMOS, above-CMOS (planar coil post-fabricated on a substrate), and around-CMOS (helical wirewound coil around substrate). We develop analytical and simulation models that have been validated in air and biological tissues by fabrications and experimental measurements. Specifically, we prototype structures that are constrained to a 4 mm $ \times$ 4 mm silicon substrate, i.e., the planar in-/above-CMOS coils have outer diameters $ 4 mm, whereas the around-CMOS coil has an inner diameter of 4 mm. The in-CMOS and above-CMOS coils have metal film thicknesses of 3- $\mu$ m aluminium and 25- $\mu$ m gold, respectively, whereas the around-CMOS coil is fabricated by winding a 25- $\mu$ m gold bonding wire around the substrate. The measured quality factors ( Q ) of the mm-scale Rx coils are 10.5 @450.3 MHz (in-CMOS), 24.61 @85 MHz (above-CMOS), and 26.23 @283 MHz (around-CMOS). Also, PTE of 2-coil links based on three types of chip-scale coils is measured in air and tissue environment to demonstrate tissue loss for bio-implants. The SAR-constrained maximum PTE measured (together with resonant frequencies, in tissue) are 1.64% @355.8 MHz (in-CMOS), 2.09% @82.9 MHz (above-CMOS), and 3.05% @318.8 MHz (around-CMOS).

This paper presents a low-power analog-to-time converter (ATC) for integrated bio-sensors. The proposed circuit facilitates the direct conversion of electrode bio-potential recordings into time-encoded digital pulses with high efficiency without prior signal amplification. This approach reduces the circuit complexity for multi-channel instrumentation systems and allows asynchronous digital control to maximize the potential power savings during sensor inactivity. A prototype fabricated using a 65-nm CMOS technology is demonstrated with measured characteristics. Experimental results show an input-referred noise figure of 3.8 $\mu V_{\mathrm{ rms}}$ for a 11-kHz signal bandwidth while dissipating 1.2 $\mu \text{W}$ from a 0.5-V supply and occupying $60\times 80\,\,\mu \text{m}\,\,^{\mathrm{ 2}}$ silicon area. This compact configuration is enabled by the proposed asynchronous readout that shapes the mismatch components arising from the multi-bit quantizer and the use of capacitive feedback.

The instrumentation systems for implantable brain–machine interfaces represent one of the most demanding applications for ultra low-power analogue-to-digital-converters (ADC) to date. To address this challenge, this paper proposes a $\Delta \Sigma $ SAR topology for very large sensor arrays that allows an exceptional reduction in silicon footprint by using a continuous time 0–2MASH topology. This configuration uses a specialized FIR window to decimate the $\Delta \Sigma $ modulator output and reject mismatch errors from the SAR quantizer, which mitigates the overhead from dynamic element matching techniques commonly used to achieve high precision. A fully differential prototype was fabricated using $0.18\,\mu $ m CMOS to demonstrate 10.8 ENOB precision with a 0.016 mm2 silicon footprint. Moreover, a 14fJ/conv figure-of-merit can be achieved, while resolving signals with the maximum input amplitude of ±1.2 Vpp sampled at 200 kS/s. The ADC topology exhibits a number of promising characteristics for both high speed and ultra low-power systems due to the reduced complexity, switching noise, sampling load, and oversampling ratio, which are critical parameters for many sensor applications.

This paper describes an on-chip interface for recovering power and providing full-duplex communication over an AC-coupled 4-wire lead between active implantable devices. The target application requires two modules to be implanted in the brain (cortex) and upper chest; connected via a subcutaneous lead. The brain implant consists of multiple identical “optrodes” that facilitate a bidirectional neural interface (electrical recording and optical stimulation), and the chest implant contains the power source (battery) and processor module. The proposed interface is integrated within each optrode ASIC allowing full-duplex and fully-differential communication based on Manchester encoding. The system features a head-to-chest uplink data rate (up to 1.6 Mbps) that is higher than that of the chest-to-head downlink (100 kbps), which is superimposed on a power carrier. On-chip power management provides an unregulated 5-V dc supply with up to 2.5-mA output current for stimulation, and two regulated voltages (3.3 and 3 V) with 60-dB power supply rejection ratio for recording and logic circuits. The 4-wire ASIC has been implemented in a 0.35-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\mu \text{m}$ \end{document} CMOS technology, occup-ying a 1.5-mm2 silicon area, and consumes a quiescent current of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$91.2~\mu \text{A}$ \end{document}. The system allows power transmission with measured efficiency of up to 66% from the chest to the brain implant. The downlink and uplink communication are successfully tested in a system with two optrodes and through a 4-wire implantable lead.

Objective: This paper describes the design, testing and use of a novel multichannel block-capable stimulator for acute neurophysiology experiments to study highly selective neural interfacing techniques. This paper demonstrates the stimulator's ability to excite and inhibit nerve activity in the rat sciatic nerve model concurrently using monophasic and biphasic nerve stimulation as well as high-frequency alternating current (HFAC). Approach: The proposed stimulator uses a Howland Current Pump circuit as the main analogue stimulator element. 4 current output channels with a common return path were implemented on printed circuit board using Commercial Off-The-Shelf components. Programmable operation is carried out by an ARM Cortex-M4 Microcontroller on the Freescale freedom development platform (K64F). Main Results: This stimulator design achieves +-10 mA of output current with +-15V of compliance and less than 6 uA of resolution using a quad-channel 12-bit external DAC, for four independently driven channels. This allows the stimulator to carry out both excitatory and inhibitory (HFAC block) stimulation. DC Output impedance is above 1 Mohm. Overall cost is less than USD 450 or GBP 350 and device size is approximately 9 cm x 6 cm x 5 cm. Significance: Experimental neurophysiology often requires significant investment in bulky equipment for specific stimulation requirements, especially when using HFAC block. Different stimulators have limited means of communicating with each other, making protocols more complicated. This device provides an effective solution for multi-channel stimulation and block of nerves, enabling studies on selective neural interfacing in acute scenarios with an affordable, portable and space-saving design for the laboratory. The stimulator can be further upgraded with additional modules to extend functionality while maintaining straightforward programming and integration of functions with one controller.

Heart rate variability (HRV), as measured by ultra-wideband (UWB) radar, enables contactless monitoring of physiological functioning in the human body. In the current study, we verified the reliability of HRV extraction from radar data, under limited transmitter power. In addition, we conducted a feasibility study of mental state classification from HRV data, measured using radar. Specifically, arctangent demodulation with calibration and low rank approximation have been used for radar signal pre-processing. An adaptive continuous wavelet filter and moving average filter were utilized for HRV extraction. For the mental state classification task, performance of support vector machine, k-nearest neighbors and random forest classifiers have been compared. The developed system has been validated on human participants, with 10 participants for HRV extraction, and three participants for the proof-of-concept mental state classification study. The results of HRV extraction demonstrate the reliability of time-domain parameter extraction from radar data. However, frequency-domain HRV parameters proved to be unreliable under low SNR. The best average overall mental state classification accuracy achieved was 82.34%, which has important implications for the feasibility of mental health monitoring using UWB radar.

Designing systems with many recording channels is a major challenge in brain-machine interfaces. Power, bandwidth, and size requirements impose tight design constraints for implementing the required processing within an acceptable latency and battery life. Moreover, the variety of brain decoding algorithms require highly versatile systems that can be rapidly adapted to execute different tasks from experiment to experiment as for example microcontrollers (MCUs). However, state of the art MCUs lack of performance and consume too much power to be used as generic platforms for neural decoding applications. To overcome the aforementioned limitations, this paper presents an MCU-based system consisting of a 64-channel event-based neural interface and a Parallel Ultra-Low-Power (PULP) platform that acquires and processes the neural activity. The flexibility of the system (called Neuro-PULP) has been demonstrated through the deployment of two applications: one compressing raw data in streaming mode for wireless transmission, and one generating the cluster and time-stamp of detected spikes, leveraging a low-power event-mode. The event-based approach, coupled with the energy efficiency of the PULP architecture, leads to a more than 4x improvement in energy efficiency with respect to state of the art systems based on FPGAs, leading to the average power consumption of $114 \mu \mathrm {W} /$channel, yet retaining the flexibility of fully programmable processor-based architectures.

The ability to acquire realtime diagnostics of brain tissue intraoperatively represents a key goal in the field of brain tumour neurosurgery. This can greatly enhance the precision, extent and effectiveness of key surgical procedures such as those performed for brain tumour resection and biopsy. To achieve this requires a miniature, handheld tool which can perform intraoperative in situ, in-vivo characterisation of different types of tissues e.g. normal brain tissue versus tumour tissue. Here we explored the feasibility and requirements of implementing a portable impedance characterisation system for brain tumour detection. We proposed and implemented a novel system based on PCB-based instrumentation using a square four-electrode microendoscopic probe. The system uses a digital-to-analogue converter to generate a multi-tone sinusoid waveform, and a floating bi-directional voltage-to-current converter to output the differential stimulation current to one pair of electrodes. The other pair of electrodes are connected to the sensing circuit based on an instrumentation amplifier. The recorded data is pre-processed by the micro-controller and then analysed on a host computer. To evaluate the system, tetrapolar impedances have been recorded from a number of different electrode configurations to sense pre-defined resistance values. The overall system consumed 143mA current, achieved 0.1% linearity and 15µV noise level, with a maximum signal bandwidth of 100kHz. Initial experimental results on tissue were carried out on a piece of rib-eye steak. Electrical impedance maps (EIM) and contour plots were then reconstructed to represent the impedance value in different tissue region.

The recent evolution of implantable medical devices from single-unit stimulators to modern implantable microsystems, has driven the need for distributed technologies, in which both the implant system and functions are partitioned across multiple active devices. This multi-module approach is made possible thanks to novel network architectures, allowing for in-body power and data communications to be performed using implantable leads. This paper discusses the challenges in implementing such interfacing system and presents a platform based on one central implant (CI) and multiple peripheral implants (PIs) using a custom 4WiCS communication protocol. This is implemented in PCB technology and tested to demonstrate intrabody communication capabilities and power transfer within the network. Measured results show CI-to-PI power delivery achieves 70 % efficiency in expected load condition, while establishing full-duplex data link with up to 4 PIs simultaneously.