Andrzej Skoczeń

Properties and application of a multichannel integrated circuit for low-artifact, patterned electrical stimulation of neural tissue

OBJECTIVE Modern multielectrode array (MEA) systems can record the neuronal activity from thousands of electrodes, but their ability to provide spatio-temporal patterns of electrical stimulation is very limited. Furthermore, the stimulus-related artifacts significantly limit the ability to record the neuronal responses to the stimulation. To address these issues, we designed a multichannel integrated circuit for a patterned MEA-based electrical stimulation and evaluated its performance in experiments with isolated mouse and rat retina. APPROACH The Stimchip includes 64 independent stimulation channels. Each channel comprises an internal digital-to-analogue converter that can be configured as a current or voltage source. The shape of the stimulation waveform is defined independently for each channel by the real-time data stream. In addition, each channel is equipped with circuitry for reduction of the stimulus artifact. MAIN RESULTS Using a high-density MEA stimulation/recording system, we effectively stimulated individual retinal ganglion cells (RGCs) and recorded the neuronal responses with minimal distortion, even on the stimulating electrodes. We independently stimulated a population of RGCs in rat retina, and using a complex spatio-temporal pattern of electrical stimulation pulses, we replicated visually evoked spiking activity of a subset of these cells with high fidelity. Significance. Compared with current state-of-the-art MEA systems, the Stimchip is able to stimulate neuronal cells with much more complex sequences of electrical pulses and with significantly reduced artifacts. This opens up new possibilities for studies of neuronal responses to electrical stimulation, both in the context of neuroscience research and in the development of neuroprosthetic devices.

Development of integrated circuits for readout of microelectrode arrays to image neuronal activity in live retinal tissue

Two multichannel Application Specific Integrated Circuits (ASICs) for extracellular recording of neuronal signals from live retinal tissues using microelectrode arrays have been developed. In this paper we discuss the requirements concerning the IC parameters and characteristics as well as present the designs and the test results. The required electronic functionality has been divided into two ASICs: PLAT-64 and NEURO-64 realized in a 0.7 /spl mu/m CMOS process. The PLAT-64 chip comprises 64 capacitors of 150 pF each, and 64 addressable and controlled DC current sources for platinization of electrodes. The NEURO-64 chip comprises 64 channels of low noise amplifiers and bandpass filters and an analog multiplexer. The low noise performance of the preamplifier has been achieved by careful selection of a CMOS process and by proper sizing and biasing of the input devices. The required lower cut-off frequency of 20 Hz has been obtained by employing a novel RC filter structure.

Using LSTM recurrent neural networks for monitoring the LHC superconducting magnets

Abstract The superconducting LHC magnets are coupled with an electronic monitoring system which records and analyzes voltage time series reflecting their performance. A currently used system is based on a range of preprogrammed triggers which launches protection procedures when a misbehavior of the magnets is detected. All the procedures used in the protection equipment were designed and implemented according to known working scenarios of the system and are updated and monitored by human operators. This paper proposes a novel approach to monitoring and fault protection of the Large Hadron Collider (LHC) superconducting magnets which employs state-of-the-art Deep Learning algorithms. Consequently, the authors of the paper decided to examine the performance of LSTM recurrent neural networks for modeling of voltage time series of the magnets. In order to address this challenging task different network architectures and hyper-parameters were used to achieve the best possible performance of the solution. The regression results were measured in terms of RMSE for different number of future steps and history length taken into account for the prediction. The best result of RMSE = 0 . 00104 was obtained for a network of 128 LSTM cells within the internal layer and 16 steps history buffer.

Fast neutron damage of silicon PIN photodiodes

A Hamamatsu Photonics photodiode S1723 was tested with respect to the fast neutron radiation. The device was irradiated with neutrons of energies in the range of 0.5 to 12 MeV from a Po-Be source. The irradiation was performed in several steps starting from the relatively low fluence of 2.5×1010 n cm−2. The following characteristics were measured: leakage current vs bias voltage, capacitance vs bias voltage and vs frequency, noise vs time constant of a quasi-Gaussian shaper and spectral density of noise. Significant changes of the leakage current and of the noise were observed at the fluence of neutrons as low as 2.5 × 1010 n cm−2.

Design of FPGA-based radiation tolerant quench detectors for LHC

The Large Hadron Collider (LHC) comprises many superconducting circuits. Most elements of these circuits require active protection. The functionality of the quench detectors was initially implemented as microcontroller based equipment. After the initial stage of the LHC operation with beams the introduction of a new type of quench detector began. This article presents briefly the main ideas and architectures applied to the design and the validation of FPGA-based quench detectors.

The model of an anomaly detector for HiLumi LHC magnets based on Recurrent Neural Networks and adaptive quantization

This paper focuses on an examination of an applicability of Recurrent Neural Network models for detecting anomalous behavior of the CERN superconducting magnets. In order to conduct the experiments, the authors designed and implemented an adaptive signal quantization algorithm and a custom GRU-based detector and developed a method for the detector parameters selection. Three different datasets were used for testing the detector. Two artificially generated datasets were used to assess the raw performance of the system whereas the 231 MB dataset composed of the signals acquired from HiLumi magnets was intended for real-life experiments and model training. Several different setups of the developed anomaly detection system were evaluated and compared with state-of-the-art OC-SVM reference model operating on the same data. The OC-SVM model was equipped with a rich set of feature extractors accounting for a range of the input signal properties. It was determined in the course of the experiments that the detector, along with its supporting design methodology, reaches F1 equal or very close to 1 for almost all test sets. Due to the profile of the data, the best_length setup of the detector turned out to perform the best among all five tested configuration schemes of the detection system. The quantization parameters have the biggest impact on the overall performance of the detector with the best values of input/output grid equal to 16 and 8, respectively. The proposed solution of the detection significantly outperformed OC-SVM-based detector in most of the cases, with much more stable performance across all the datasets.Abstract This paper focuses on an examination of an applicability of Recurrent Neural Network models for detecting anomalous behavior of the CERN superconducting magnets. In order to conduct the experiments, the authors designed and implemented an adaptive signal quantization algorithm and a custom Gated Recurrent Unit-based detector and developed a method for the detector parameters selection. Three different datasets were used for testing the detector. Two artificially generated datasets were used to assess the raw performance of the system whereas the dataset intended for real-life experiments and model training was composed of the signals acquired from a new type of magnet, to be used during High-Luminosity Large Hadron Collider project. Several different setups of the developed anomaly detection system were evaluated and compared with state-of-the-art One Class Support Vector Machine (OC-SVM) reference model operating on the same data. The OC-SVM model was equipped with a rich set of feature extractors accounting for a range of the input signal properties. It was determined in the course of the experiments that the detector, along with its supporting design methodology, reaches F1 equal or very close to 1 for almost all test sets. Due to the profile of the data, the setup with the lowest maximum false anomaly length of the detector turned out to perform the best among all five tested configuration schemes of the detection system. The quantization parameters have the biggest impact on the overall performance of the detector with the best values of input/output grid equal to 16 and 8, respectively. The proposed solution of the detection significantly outperformed OC-SVM-based detector in most of the cases, with much more stable performance across all the datasets.

Modular ASIC-based system for large-scale electrical stimulation and recording of brain activity in behaving animals

We present a design and first test results of a novel microelectronic system for large-scale electrical stimulation and recording of brain activity in the behaving animals. The system is based on a new 64-channel ASIC being designed in the AMS 0.35 μm CMOS process and is compatible with all modern multielectrode silicon probes for brain research. It allows to record neuronal activity at up to 512 electrodes and to generate complex, arbitrarily defined sequences of stimulation pulses across all the electrodes. Modular system design makes it possible to communicate simultaneously with neuronal circuits in several brain areas, providing direct information about functional connectivities between different brain regions.