A 16-Channel Fully Configurable Neural SoC With 1.52 µW/Ch Signal Acquisition, 2.79 µW/Ch Real-Time Spike Classifier, and 1.79 TOPS/W Deep Neural Network Accelerator in 22 nm FDSOI.

With the advent of high-density micro-electrodes arrays, developing neural probes satisfying the real-time and stringent power-efficiency requirements becomes more challenging. A smart neural probe is an essential device in future neuroscientific research and medical applications. To realize such devices, we present a 22 nm FDSOI SoC with complex on-chip real-time data processing and training for neural signal analysis. It consists of a digitally-assisted 16-channel analog front-end with 1.52 µW/Ch, dedicated bio-processing accelerators for spike detection and classification with 2.79 µW/Ch, and a 125 MHz RISC-V CPU, utilizing adaptive body biasing at 0.5 V with a supporting 1.79 TOPS/W MAC array. The proposed SoC shows a proof-of-concept of how to realize a high-level integration of various on-chip accelerators to satisfy the neural probe requirements for modern applications.

Hardware Acceleration of EEG-Based Emotion Classification Systems: A Comprehensive Survey.

Recent years have witnessed a growing interest in EEG-based wearable classifiers of emotions, which could enable the real-time monitoring of patients suffering from neurological disorders such as Amyotrophic Lateral Sclerosis (ALS), Autism Spectrum Disorder (ASD), or Alzheimer's. The hope is that such wearable emotion classifiers would facilitate the patients' social integration and lead to improved healthcare outcomes for them and their loved ones. Yet in spite of their direct relevance to neuro-medicine, the hardware platforms for emotion classification have yet to fill up some important gaps in their various approaches to emotion classification in a healthcare context. In this paper, we present the first hardware-focused critical review of EEG-based wearable classifiers of emotions and survey their implementation perspectives, their algorithmic foundations, and their feature extraction methodologies. We further provide a neuroscience-based analysis of current hardware accelerators of emotion classifiers and use it to map out several research opportunities, including multi-modal hardware platforms, accelerators with tightly-coupled cores operating robustly in the near/supra-threshold region, and pre-processing libraries for universal EEG-based datasets.

Efficient Reward-Based Structural Plasticity on a SpiNNaker 2 Prototype.

Advances in neuroscience uncover the mechanisms employed by the brain to efficiently solve complex learning tasks with very limited resources. However, the efficiency is often lost when one tries to port these findings to a silicon substrate, since brain-inspired algorithms often make extensive use of complex functions, such as random number generators, that are expensive to compute on standard general purpose hardware. The prototype chip of the second generation SpiNNaker system is designed to overcome this problem. Low-power advanced RISC machine (ARM) processors equipped with a random number generator and an exponential function accelerator enable the efficient execution of brain-inspired algorithms. We implement the recently introduced reward-based synaptic sampling model that employs structural plasticity to learn a function or task. The numerical simulation of the model requires to update the synapse variables in each time step including an explorative random term. To the best of our knowledge, this is the most complex synapse model implemented so far on the SpiNNaker system. By making efficient use of the hardware accelerators and numerical optimizations, the computation time of one plasticity update is reduced by a factor of 2. This, combined with fitting the model into to the local static random access memory (SRAM), leads to 62% energy reduction compared to the case without accelerators and the use of external dynamic random access memory (DRAM). The model implementation is integrated into the SpiNNaker software framework allowing for scalability onto larger systems. The hardware-software system presented in this paper paves the way for power-efficient mobile and biomedical applications with biologically plausible brain-inspired algorithms.

Memory-Efficient Deep Learning on a SpiNNaker 2 Prototype.

The memory requirement of deep learning algorithms is considered incompatible with the memory restriction of energy-efficient hardware. A low memory footprint can be achieved by pruning obsolete connections or reducing the precision of connection strengths after the network has been trained. Yet, these techniques are not applicable to the case when neural networks have to be trained directly on hardware due to the hard memory constraints. Deep Rewiring (DEEP R) is a training algorithm which continuously rewires the network while preserving very sparse connectivity all along the training procedure. We apply DEEP R to a deep neural network implementation on a prototype chip of the 2nd generation SpiNNaker system. The local memory of a single core on this chip is limited to 64 KB and a deep network architecture is trained entirely within this constraint without the use of external memory. Throughout training, the proportion of active connections is limited to 1.3%. On the handwritten digits dataset MNIST, this extremely sparse network achieves 96.6% classification accuracy at convergence. Utilizing the multi-processor feature of the SpiNNaker system, we found very good scaling in terms of computation time, per-core memory consumption, and energy constraints. When compared to a X86 CPU implementation, neural network training on the SpiNNaker 2 prototype improves power and energy consumption by two orders of magnitude.

A Biological-Realtime Neuromorphic System in 28 nm CMOS Using Low-Leakage Switched Capacitor Circuits.

A switched-capacitor (SC) neuromorphic system for closed-loop neural coupling in 28 nm CMOS is presented, occupying 600 um by 600 um. It offers 128 input channels (i.e., presynaptic terminals), 8192 synapses and 64 output channels (i.e., neurons). Biologically realistic neuron and synapse dynamics are achieved via a faithful translation of the behavioural equations to SC circuits. As leakage currents significantly affect circuit behaviour at this technology node, dedicated compensation techniques are employed to achieve biological-realtime operation, with faithful reproduction of time constants of several 100 ms at room temperature. Power draw of the overall system is 1.9 mW.

Complementing the characterization of in vivo generated N-glucuronic acid conjugates of stanozolol by collision cross section computation and analysis.

Detailed structural information on metabolites serving as target analytes in clinical, forensic, and sports drug testing programmes is of paramount importance to ensure unequivocal test results. In the present study, the utility of collision cross section (CCS) analysis by travelling wave ion mobility measurements to support drug metabolite characterization efforts was tested concerning recently identified glucuronic acid conjugates of the anabolic-androgenic steroid stanozolol. Employing travelling-wave ion mobility spectrometry/quadrupole-time-of-flight mass spectrometry, drift times of five synthetically derived and fully characterized steroid glucuronides were measured and subsequently correlated to respective CCSs as obtained in silico to form an analyte-tailored calibration curve. The CCSs were calculated by equilibrium structure minimization (density functional theory) using the programmes ORCA with the data set B3LYP/6-31G and MOBCAL utilizing the trajectory method (TM) with nitrogen as drift gas. Under identical experimental conditions, synthesized and/or urinary stanozolol-N and O-glucuronides were analyzed to provide complementary information on the location of glucuronidation. Finally, the obtained data were compared to CCS results generated by the system's internal algorithm based on a calibration employing a polyalanine analyte mixture. The CCSs ΩN2 calculated for the five steroid glucuronide calibrants were found between 180 and 208 Å(2) , thus largely covering the observed and computed CCSs for stanozolol-N1'-, stanozolol-N2'-, and stanozolol-O-glucuronide found at values between 195.1 and 212.4 Å(2) . The obtained data corroborated the earlier suggested N- and O-glucuronidation of stanozolol, and demonstrate the exploit of ion mobility and CCS computation in structure characterization of phase-II metabolic products; however, despite reproducibly measurable differences in ion mobility of stanozolol-N1'-, N2'-, and O-glucuronides, the discriminatory power of the chosen CCS computation algorithm was found to be not appropriate to allow for accurate assignments of the two N-conjugated structures. Using polyalanine-based calibrations, significantly different absolute values were obtained for all CCSs, but due to a constant offset of approximately 45 Å(2) an excellent correlation (R(2) = 0.9997) between both approaches was observed. This suggests a substantially accelerated protocol when patterns of computed and polyalanine-based experimental data can be used for structure elucidations instead of creating individual analyte-specific calibration curves.

Rapid high resolution genotyping of Francisella tularensis by whole genome sequence comparison of annotated genes ("MLST+").

The zoonotic disease tularemia is caused by the bacterium Francisella tularensis. This pathogen is considered as a category A select agent with potential to be misused in bioterrorism. Molecular typing based on DNA-sequence like canSNP-typing or MLVA has become the accepted standard for this organism. Due to the organism's highly clonal nature, the current typing methods have reached their limit of discrimination for classifying closely related subpopulations within the subspecies F. tularensis ssp. holarctica. We introduce a new gene-by-gene approach, MLST+, based on whole genome data of 15 sequenced F. tularensis ssp. holarctica strains and apply this approach to investigate an epidemic of lethal tularemia among non-human primates in two animal facilities in Germany. Due to the high resolution of MLST+ we are able to demonstrate that three independent clones of this highly infectious pathogen were responsible for these spatially and temporally restricted outbreaks.

Reducing the computational footprint for real-time BCPNN learning.

The implementation of synaptic plasticity in neural simulation or neuromorphic hardware is usually very resource-intensive, often requiring a compromise between efficiency and flexibility. A versatile, but computationally-expensive plasticity mechanism is provided by the Bayesian Confidence Propagation Neural Network (BCPNN) paradigm. Building upon Bayesian statistics, and having clear links to biological plasticity processes, the BCPNN learning rule has been applied in many fields, ranging from data classification, associative memory, reward-based learning, probabilistic inference to cortical attractor memory networks. In the spike-based version of this learning rule the pre-, postsynaptic and coincident activity is traced in three low-pass-filtering stages, requiring a total of eight state variables, whose dynamics are typically simulated with the fixed step size Euler method. We derive analytic solutions allowing an efficient event-driven implementation of this learning rule. Further speedup is achieved by first rewriting the model which reduces the number of basic arithmetic operations per update to one half, and second by using look-up tables for the frequently calculated exponential decay. Ultimately, in a typical use case, the simulation using our approach is more than one order of magnitude faster than with the fixed step size Euler method. Aiming for a small memory footprint per BCPNN synapse, we also evaluate the use of fixed-point numbers for the state variables, and assess the number of bits required to achieve same or better accuracy than with the conventional explicit Euler method. All of this will allow a real-time simulation of a reduced cortex model based on BCPNN in high performance computing. More important, with the analytic solution at hand and due to the reduced memory bandwidth, the learning rule can be efficiently implemented in dedicated or existing digital neuromorphic hardware.

Switched-capacitor realization of presynaptic short-term-plasticity and stop-learning synapses in 28 nm CMOS.

Synaptic dynamics, such as long- and short-term plasticity, play an important role in the complexity and biological realism achievable when running neural networks on a neuromorphic IC. For example, they endow the IC with an ability to adapt and learn from its environment. In order to achieve the millisecond to second time constants required for these synaptic dynamics, analog subthreshold circuits are usually employed. However, due to process variation and leakage problems, it is almost impossible to port these types of circuits to modern sub-100nm technologies. In contrast, we present a neuromorphic system in a 28 nm CMOS process that employs switched capacitor (SC) circuits to implement 128 short term plasticity presynapses as well as 8192 stop-learning synapses. The neuromorphic system consumes an area of 0.36 mm(2) and runs at a power consumption of 1.9 mW. The circuit makes use of a technique for minimizing leakage effects allowing for real-time operation with time constants up to several seconds. Since we rely on SC techniques for all calculations, the system is composed of only generic mixed-signal building blocks. These generic building blocks make the system easy to port between technologies and the large digital circuit part inherent in an SC system benefits fully from technology scaling.

Mass spectrometric studies on the in vivo metabolism and excretion of SIRT1 activating drugs in rat urine, dried blood spots, and plasma samples for doping control purposes.

The NAD(+) depending enzyme SIRT1 regulates the mitochondrial biogenesis, fat and glucose metabolism through catalyzing the deacetylation of several metabolism-related protein-substrates. Recently, synthetic activators of SIRT1 referred to as STACs (Sirtuin activating compounds, e.g. SRT2104) were identified and tested in clinical studies for the treatment of aging-related diseases such as type 2 diabetes, Alzheimer's and obesity. Although the mechanism of SIRT1 activation by small molecules has caused considerable controversy, STACs demonstrated a significant performance enhancement in mice experiments including an improvement of endurance, muscle strength, and locomotor behavior. Due to their potential to increase exercise tolerance in healthy individuals, SIRT1 activators are currently being monitored by anti-doping authorities. In the present study, the in vivo metabolic clearance of three SIRT1 activators was investigated in rats by the collection of urine, DBS (dried blood spots) and plasma samples following a single oral administration. The resulting metabolic products were studied by positive electrospray ionization - (tandem) mass spectrometry and confirmed by the comparison with in vitro generated metabolites using human and rat liver microsomal preparations. Subsequently, a screening procedure for five SIRT1 activators and the metabolite M1-SRT1720 in DBS specimens was developed. Liquid-liquid-extraction and liquid chromatography/tandem mass spectrometry was employed based on diagnostic ion transitions recorded in multiple reaction monitoring mode and two deuterated internal standards namely d8-SRT1720 and d8-M1-SRT1720 were utilized. The doping control assay was characterized with regard to specificity, limit of detection (10-50ng/ml), recovery (65-83%) and imprecision (7-20%) and ion suppression/enhancement effects (<10%), demonstrating its fitness-for-purpose for sports drug testing applications.

Mass spectrometric studies on the in vitro generated metabolites of SIRT1 activating drugs for doping control purposes.

The enzyme SIRT1 is a metabolic key regulator in mitochondrial biogenesis, fat and glucose metabolism. Its activation through pharmaceutical SIRT1 activators such as SRT2104 results in an increased deacetylation of substrates representing important targets for the treatment of metabolic diseases. Moreover, SRT1720 was found to enhance the physical performance of mice. As SIRT1 activators might therefore be relevant in a doping control context, metabolism studies of target substances need be conducted in order to develop a detection assay for SIRT1 activators in urine. In the present study, the in vitro metabolism of five SIRT1 activators was investigated using human liver microsomes. The mass spectrometric behavior of the resulting metabolites following positive electrospray ionization and collision-induced dissociation was elucidated by high-resolution/high-accuracy (tandem) mass spectrometry, and confirmation of the structure of a major metabolite of SRT1720 was accomplished by chemical synthesis. Subsequently, a screening procedure for urine samples was developed employing liquid-liquid-extraction and liquid chromatography/tandem mass spectrometry based on diagnostic ion transitions recorded in multiple reaction monitoring mode and the use of d8-SRT1720 as deuterated internal standard. The method was validated with regard to specificity, sensitivity (limit of detection 0.5 ng/ml), recovery (88-99%) and imprecision (7-18%) as well as ion suppression/enhancement effects (<10%), demonstrating its fitness-for-purpose for sports drug testing applications.

Fragmentation studies of SIRT1-activating drugs and their detection in human plasma for doping control purposes.

RATIONALE: The efficiency of Sirtuin1, a major target for the treatment of various metabolic disorders such as inflammation and type 2 diabetes mellitus, can be modulated via low molecular mass SIRT1 activators (e.g. resveratrol, SRT1720, and SRT2104).The administration of such compounds results in increased deacetylation of substrates including p53, FOXO1, and PGC1alpha, potentially leading to an improved physical performance. Consequently, proactive and preventive anti-doping measures are required and an assay dedicated to serum and plasma was desirable. METHODS: Model substances of emerging SIRT1 drug candidates were obtained and synthesized and their mass spectrometric behavior following positive or negative electrospray ionization and collision-induced dissociation was elucidated using low and high resolution/high accuracy (tandem) mass spectrometry. Subsequently, a screening and confirmation procedure necessitating 100 µL of plasma was established employing liquid chromatography/tandem mass spectrometry (LC/MS/MS) based on diagnostic ion transitions recorded in multiple reaction monitoring mode. Sample preparation consisted of the addition of two deuterated internal standards (D(8)-SRT1720 and D(4)-resveratrol) to the plasma specimen and subsequent protein precipitation. RESULTS: Characteristic product ions indicative of the core structures of the model analytes were characterized and utilized for the development of a multi-analyte LC/MS/MS detection method applicable to sports drug testing programs. The doping control assay was validated with regard to specificity, limits of detection (0.1-1 ng/mL), recoveries (90-98%), intraday and interday precisions (2-18%), and ion suppression/enhancement effects. CONCLUSIONS: The fragmentation pathways of SRT1720 and 4 SIRT1 activator models based on a common thiazole-imidazole nucleus as well as two different complementary activators (SIRT1 activator 3 and CAY10602), comprising a quinoxaline core, were studied. The resulting information was used to establish and validate a sports drug testing methodology relevant for an efficient and timely anti-doping procedure, targeting a new class of emerging therapeutics possessing significant potential for misuse in elite and amateur sport.

Structure elucidation of the diagnostic product ion at m/z 97 derived from androst-4-en-3-one-based steroids by ESI-CID and IRMPD spectroscopy.

Structure elucidation of steroids by mass spectrometry has been of great importance to various analytical arenas and numerous studies were conducted to provide evidence for the composition and origin of (tandem) mass spectrometry-derived product ions used to characterize and identify steroidal substances. The common product ion at m/z 97 generated from androst-4-ene-3-one analogs has been subject of various studies, including stable isotope-labeling and (high resolution/high accuracy) tandem mass spectrometry, but its gas-phase structure has never been confirmed. Using high resolution/high accuracy mass spectrometry and low resolution tandem mass spectrometry, density functional theory (DFT) calculation, and infrared multiple photon dissociation (IRMPD) spectroscopy employing a free electron laser, the structure of m/z 97 derived from testosterone was assigned to protonated 3-methyl-2-cyclopenten-1-one. This ion was identified in a set of six cyclic C(6)H(9)O(+) isomers as computed at the B3LYP/6-311++G(2d,2p) level of theory (protonated 3-methyl-2-cyclopenten-1-one, 2-methyl-2-cyclopenten-1-one and 2-cyclohexen-1-one). Product ions of m/z 97 obtained from MS(2) and MS(3) experiments of protonated 3-methyl-2-cyclopenten-1-one, 2-methyl-2-cyclopenten-1-one, 2-cyclohexen-1-one, and testosterone corroborated the suggested gas-phase ion structure, which was eventually substantiated by IRMPD spectroscopy yielding a spectrum that convincingly matched the predicted counterpart. Finally, the dissociation pathway of the protonated molecule of testosterone to m/z 97 was revisited and an alternative pathway was suggested that considers the exclusion of C-10 along with the inclusion of C-5, which was experimentally demonstrated with stable isotope labeling.

VLSI Implementation of a 2.8 Gevent/s Packet-Based AER Interface with Routing and Event Sorting Functionality.

State-of-the-art large-scale neuromorphic systems require sophisticated spike event communication between units of the neural network. We present a high-speed communication infrastructure for a waferscale neuromorphic system, based on application-specific neuromorphic communication ICs in an field programmable gate arrays (FPGA)-maintained environment. The ICs implement configurable axonal delays, as required for certain types of dynamic processing or for emulating spike-based learning among distant cortical areas. Measurements are presented which show the efficacy of these delays in influencing behavior of neuromorphic benchmarks. The specialized, dedicated address-event-representation communication in most current systems requires separate, low-bandwidth configuration channels. In contrast, the configuration of the waferscale neuromorphic system is also handled by the digital packet-based pulse channel, which transmits configuration data at the full bandwidth otherwise used for pulse transmission. The overall so-called pulse communication subgroup (ICs and FPGA) delivers a factor 25-50 more event transmission rate than other current neuromorphic communication infrastructures.

Determination of growth hormone releasing peptides (GHRP) and their major metabolites in human urine for doping controls by means of liquid chromatography mass spectrometry.

A family of small peptides has reached the focus of doping controls representing a comparably new strategy for cheating sportsmen. These growth hormone releasing peptides (GHRP) are orally active and induce an increased production of endogenous growth hormone (GH). While the established test for exogenous GH fails, the misuse of these prohibited substances remains unrecognized. The present study provides data for the efficient extraction of a variety of known drug candidates (GHRP-1, GHRP-2, GHRP-4, GHRP-5, GHRP-6, alexamorelin, ipamorelin, and hexarelin) from human urine with subsequent mass spectrometric detection after liquid chromatographic separation. The used method potentially enables the retrospective evaluation of the acquired data for unknown metabolites by means of a non-targeted approach with high-resolution/high-accuracy full-scan mass spectrometry with additional higher collision energy dissociation experiments. This is of great importance due to the currently unknown metabolism of most of the targets and, thus, the method is focused on the intact peptidic drugs. Only the already characterised major metabolite of GHRP-2 (D-Ala-D-2-naphthylAla-L-Ala, as well as its stable isotope-labelled analogue) was synthesised and implemented in the detection assay. Method validation for qualitative purpose was performed with respect to specificity, precision (<20%), intermediate precision (<20%), recovery (47-95%), limit of detection (0.2-1 ng/mL), linearity, ion suppression and stability. Two stable isotope-labelled internal standards were used (deuterium-labelled GHRP-4 and GHRP-2 metabolite). The proof-of-principle was obtained by the analysis of excretion study urine samples obtained from a single oral administration of 10 mg of GHRP-2. Here, the known metabolite was detectable over 20 h after administration while the intact drug was not observed.