Edge computing on TPU for brain implant signal analysis.

The ever-increasing number of recording sites of silicon-based probes imposes a great challenge for detecting and evaluating single-unit activities in an accurate and efficient manner. Currently separate solutions are available for high precision offline evaluation and separate solutions for embedded systems where computational resources are more limited. We propose a deep learning-based spike sorting system, that utilizes both unsupervised and supervised paradigms to learn a general feature embedding space and detect neural activity in raw data as well as predict the feature vectors for sorting. The unsupervised component uses contrastive learning to extract features from individual waveforms, while the supervised component is based on the MobileNetV2 architecture. One of the key advantages of our system is that it can be trained on multiple, diverse datasets simultaneously, resulting in greater generalizability than previous deep learning-based models. We demonstrate that the proposed model does not only reaches the accuracy of current state-of-art offline spike sorting methods but has the unique potential to run on edge Tensor Processing Units (TPUs), specialized chips designed for artificial intelligence and edge computing. We compare our model performance with state of art solutions on paired datasets as well as on hybrid recordings as well. The herein demonstrated system paves the way to the integration of deep learning-based spike sorting algorithms into wearable electronic devices, which will be a crucial element of high-end brain-computer interfaces.

From End to End: Gaining, Sorting, and Employing High-Density Neural Single Unit Recordings.

The meaning behind neural single unit activity has constantly been a challenge, so it will persist in the foreseeable future. As one of the most sourced strategies, detecting neural activity in high-resolution neural sensor recordings and then attributing them to their corresponding source neurons correctly, namely the process of spike sorting, has been prevailing so far. Support from ever-improving recording techniques and sophisticated algorithms for extracting worthwhile information and abundance in clustering procedures turned spike sorting into an indispensable tool in electrophysiological analysis. This review attempts to illustrate that in all stages of spike sorting algorithms, the past 5 years innovations' brought about concepts, results, and questions worth sharing with even the non-expert user community. By thoroughly inspecting latest innovations in the field of neural sensors, recording procedures, and various spike sorting strategies, a skeletonization of relevant knowledge lays here, with an initiative to get one step closer to the original objective: deciphering and building in the sense of neural transcript.

PlatypOUs-A Mobile Robot Platform and Demonstration Tool Supporting STEM Education.

Given the rising popularity of robotics, student-driven robot development projects are playing a key role in attracting more people towards engineering and science studies. This article presents the early development process of an open-source mobile robot platform-named PlatypOUs-which can be remotely controlled via an electromyography (EMG) appliance using the MindRove brain-computer interface (BCI) headset as a sensor for the purpose of signal acquisition. The gathered bio-signals are classified by a Support Vector Machine (SVM) whose results are translated into motion commands for the mobile platform. Along with the physical mobile robot platform, a virtual environment was implemented using Gazebo (an open-source 3D robotic simulator) inside the Robot Operating System (ROS) framework, which has the same capabilities as the real-world device. This can be used for development and test purposes. The main goal of the PlatypOUs project is to create a tool for STEM education and extracurricular activities, particularly laboratory practices and demonstrations. With the physical robot, the aim is to improve awareness of STEM outside and beyond the scope of regular education programmes. It implies several disciplines, including system design, control engineering, mobile robotics and machine learning with several application aspects in each. Using the PlatypOUs platform and the simulator provides students and self-learners with a firsthand exercise, and teaches them to deal with complex engineering problems in a professional, yet intriguing way.

ELVISort: encoding latent variables for instant sorting, an artificial intelligence-based end-to-end solution.

Objective.The growing number of recording sites of silicon-based probes means that an increasing amount of neural cell activities can be recorded simultaneously, facilitating the investigation of underlying complex neural dynamics. In order to overcome the challenges generated by the increasing number of channels, highly automated signal processing tools are needed. Our goal was to build a spike sorting model that can perform as well as offline solutions while maintaining high efficiency, enabling high-performance online sorting.Approach.In this paper we present ELVISort, a deep learning method that combines the detection and clustering of different action potentials in an end-to-end fashion.Main results.The performance of ELVISort is comparable with other spike sorting methods that use manual or semi-manual techniques, while exceeding the methods which use an automatic approach: ELVISort has been tested on three independent datasets and yielded average F1scores of 0.96, 0.82 and 0.81, which comparable with the results of state-of-the-art algorithms on the same data. We show that despite the good performance, ELVISort is capable to process data in real-time: the time it needs to execute the necessary computations for a sample of given length is only 1/15.71 of its actual duration (i.e. the sampling time multiplied by the number of the sampling points).Significance.ELVISort, because of its end-to-end nature, can exploit the massively parallel processing capabilities of GPUs via deep learning frameworks by processing multiple batches in parallel, with the potential to be used on other cutting-edge AI-specific hardware such as TPUs, enabling the development of integrated, portable and real-time spike sorting systems with similar performance to offline sorters.

The neural tissue around SU-8 implants: A quantitative in vivo biocompatibility study.

The use of SU-8 material in the production of neural sensors has grown recently. Despite its widespread application, a detailed systematic quantitative analysis concerning its biocompatibility in the central nervous system is lacking. In this immunohistochemical study, we quantified the neuronal preservation and the severity of astrogliosis around SU-8 devices implanted in the neocortex of rats, after a 2 months survival. We found that the density of neurons significantly decreased up to a distance of 20 µm from the implant, with an averaged density decrease to 24 ±â€¯28% of the control. At 20 to 40 µm distance from the implant, the majority of the neurons was preserved (74 ±â€¯39% of the control) and starting from 40 µm distance from the implant, the neuron density was control-like. The density of synaptic contacts - examined at the electron microscopic level - decreased in the close vicinity of the implant, but it recovered to the control level as close as 24 µm from the implant track. The intensity of the astroglial staining significantly increased compared to the control region, up to 560 µm and 480 µm distance from the track in the superficial and deep layers of the neocortex, respectively. Electron microscopic examination revealed that the thickness of the glial scar was around 5-10 µm thin, and the ratio of glial processes in the neuropil was not more than 16% up to a distance of 12 µm from the implant. Our data suggest that neuronal survival is affected only in a very small area around the implant. The glial scar surrounding the implant is thin, and the presence of glial elements is low in the neuropil, although the signs of astrogliosis could be observed up to about 500 µm from the track. Subsequently, the biocompatibility of the SU-8 material is high. Due to its low cost fabrication and more flexible nature, SU-8 based devices may offer a promising approach to experimental and clinical applications in the future.

Spike detection and sorting with deep learning.

OBJECTIVE: The extraction and identification of single-unit activities in intracortically recorded electric signals have a key role in basic neuroscience, but also in applied fields, like in the development of high-accuracy brain-computer interfaces. The purpose of this paper is to present our current results on the detection, classification and prediction of neural activities based on multichannel action potential recordings. APPROACH: Throughout our investigations, a deep learning approach utilizing convolutional neural networks and a combination of recurrent and convolutional neural networks was applied, with the latter used in case of spike detection and the former used for cases of sorting and predicting spiking activities. MAIN RESULTS: In our experience, the algorithms applied prove to be useful in accomplishing the tasks mentioned above: our detector could reach an average recall of 69%, while we achieved an average accuracy of 89% in classifying activities produced by more than 20 distinct neurons. SIGNIFICANCE: Our findings support the concept of creating real-time, high-accuracy action potential based BCIs in the future, providing a flexible and robust algorithmic background for further development.

Correction: Method for spike detection from microelectrode array recordings contaminated by artifacts of simultaneous two-photon imaging.

[This corrects the article DOI: 10.1371/journal.pone.0221510.].

Method for spike detection from microelectrode array recordings contaminated by artifacts of simultaneous two-photon imaging.

The simultaneous utilization of electrophysiological recordings and two-photon imaging allows the observation of neural activity in a high temporal and spatial resolution at the same time. The three dimensional monitoring of morphological features near the microelectrode array makes the observation more precise and complex. In vitro experiments were performed on mice neocortical slices expressing the GCaMP6 genetically encoded calcium indicator for monitoring the neural activity with two-photon microscopy around the implanted microelectrodes. A special filtering algorithm was used for data analysis to eliminate the artefacts caused by the imaging laser. Utilization of a special filtering algorithm allowed us to detect and sort single unit activities from simultaneous two-photon imaging and electrophysiological measurement.

Slow insertion of silicon probes improves the quality of acute neuronal recordings.

Neural probes designed for extracellular recording of brain electrical activity are traditionally implanted with an insertion speed between 1 µm/s and 1 mm/s into the brain tissue. Although the physical effects of insertion speed on the tissue are well studied, there is a lack of research investigating how the quality of the acquired electrophysiological signal depends on the speed of probe insertion. In this study, we used four different insertion speeds (0.002 mm/s, 0.02 mm/s, 0.1 mm/s, 1 mm/s) to implant high-density silicon probes into deep layers of the somatosensory cortex of ketamine/xylazine anesthetized rats. After implantation, various qualitative and quantitative properties of the recorded cortical activity were compared across different speeds in an acute manner. Our results demonstrate that after the slowest insertion both the signal-to-noise ratio and the number of separable single units were significantly higher compared with those measured after inserting probes at faster speeds. Furthermore, the amplitude of recorded spikes as well as the quality of single unit clusters showed similar speed-dependent differences. Post hoc quantification of the neuronal density around the probe track showed a significantly higher number of NeuN-labelled cells after the slowest insertion compared with the fastest insertion. Our findings suggest that advancing rigid probes slowly (~1 µm/s) into the brain tissue might result in less tissue damage, and thus in neuronal recordings of improved quality compared with measurements obtained after inserting probes with higher speeds.

Neurobiochemical changes in the vicinity of a nanostructured neural implant.

Neural interface technologies including recording and stimulation electrodes are currently in the early phase of clinical trials aiming to help patients with spinal cord injuries, degenerative disorders, strokes interrupting descending motor pathways, or limb amputations. Their lifetime is of key importance; however, it is limited by the foreign body response of the tissue causing the loss of neurons and a reactive astrogliosis around the implant surface. Improving the biocompatibility of implant surfaces, especially promoting neuronal attachment and regeneration is therefore essential. In our work, bioactive properties of implanted black polySi nanostructured surfaces (520-800 nm long nanopillars with a diameter of 150-200 nm) were investigated and compared to microstructured Si surfaces in eight-week-long in vivo experiments. Glial encapsulation and local neuronal cell loss were characterised using GFAP and NeuN immunostaining respectively, followed by systematic image analysis. Regarding the severity of gliosis, no significant difference was observed in the vicinity of the different implant surfaces, however, the number of surviving neurons close to the nanostructured surface was higher than that of the microstructured ones. Our results imply that the functionality of implanted microelectrodes covered by Si nanopillars may lead to improved long-term recordings.

A Multimodal, SU-8 - Platinum - Polyimide Microelectrode Array for Chronic In Vivo Neurophysiology.

Utilization of polymers as insulator and bulk materials of microelectrode arrays (MEAs) makes the realization of flexible, biocompatible sensors possible, which are suitable for various neurophysiological experiments such as in vivo detection of local field potential changes on the surface of the neocortex or unit activities within the brain tissue. In this paper the microfabrication of a novel, all-flexible, polymer-based MEA is presented. The device consists of a three dimensional sensor configuration with an implantable depth electrode array and brain surface electrodes, allowing the recording of electrocorticographic (ECoG) signals with laminar ones, simultaneously. In vivo recordings were performed in anesthetized rat brain to test the functionality of the device under both acute and chronic conditions. The ECoG electrodes recorded slow-wave thalamocortical oscillations, while the implanted component provided high quality depth recordings. The implants remained viable for detecting action potentials of individual neurons for at least 15 weeks.

Durability of high surface area platinum deposits on microelectrode arrays for acute neural recordings.

The durability of high surface area platinum electrodes during acute intracerebral measurements was investigated. Electrode sites with extremely rough surfaces were realized using electrochemical deposition of platinum onto silicon-based microelectrode arrays from a lead-free platinizing solution. The close to 1000-fold increase in effective surface area lowered impedance, its absolute value at 1 kHz became about 7 and 18 % of the original Pt electrodes in vitro and in vivo, respectively. 24-channel probes were subjected to 12 recording sessions, during which they were implanted into the cerebrum of rats. Our results showed that although on the average the effective surface area of the platinized sites decreased, it remained more than two orders of magnitude higher than the average effective surface area of the original, sputtered thin-film platinum electrodes. Sites with electrochemical deposits proved to be superior, e.g. they provided less thermal and 50 Hz noise, even after 12 penetrations into the intact rat brain.