Agnese Grison
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Agnese Grison has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Dario Farina
Department of Bioengineering, Imperial College London, London, UK
7 shared publications - Aritra Kundu
Department of Bioengineering, Imperial College London
4 shared publications - Jaime Ibáñez
BSICoS group, I3A Institute and IIS Aragon, University of Zaragoza
3 shared publications - Silvia Muceli
Department of Bioengineering, Imperial College London, U.K.
3 shared publications - Alexander Kenneth Clarke
Department of Bioengineering, Imperial College London, London, U.K.
2 shared publications - Ciara Gibbs
Imperial College London
2 shared publications
Latest Publications
- Open Access
Intuitive, multidimensional motor unit control after paralysing spinal cord injury using non-invasive recordings
Spinal cord injury (SCI) results in profound motor impairment for approximately 20 million people worldwide. Regaining hand use is one of their highest priorities. Interestingly, even severely affected individuals can still generate some level of muscle activity (EMG) in their paralysed muscles when attempting to perform a movement. Here we asked whether participants with no hand function due to a cervical SCI could still use their spared inputs to volitionally control the activity of motor units from paralysed muscles, which we detected using high-density surface EMG recordings. Participants could readily control those motor units to accurately perform three different one-dimensional tasks that imposed different control requirements, with their performance plateauing after a few days of practice. Moreover, this accurate control readily transferred to a two-dimensional cursor navigation task and allowed participants to play a series of one-dimensional ('Pong'), and two-dimensional ('Snake') games, as well as navigate a virtual wheelchair using three independent control signals. This robust performance was achieved despite the presence of spasticity, and even in one participant who had a motor complete injury. These findings demonstrate the promise of non-invasive motor unit recordings for restoring hand function, as well as a potential tool for neurorehabilitation after paralysis.
Open AccessIntuitive, multidimensional motor unit control after paralysing spinal cord injury using non-invasive recordings
Spinal cord injury (SCI) results in profound motor impairment for approximately 20 million people worldwide. Regaining hand use is one of their highest priorities. Interestingly, even severely affected individuals can still generate some level of muscle activity (EMG) in their paralysed muscles when attempting to perform a movement. Here we asked whether participants with no hand function due to a cervical SCI could still use their spared inputs to volitionally control the activity of motor units from paralysed muscles, which we detected using high-density surface EMG recordings. Participants could readily control those motor units to accurately perform three different one-dimensional tasks that imposed different control requirements, with their performance plateauing after a few days of practice. Moreover, this accurate control readily transferred to a two-dimensional cursor navigation task and allowed participants to play a series of one-dimensional ('Pong'), and two-dimensional ('Snake') games, as well as navigate a virtual wheelchair using three independent control signals. This robust performance was achieved despite the presence of spasticity, and even in one participant who had a motor complete injury. These findings demonstrate the promise of non-invasive motor unit recordings for restoring hand function, as well as a potential tool for neurorehabilitation after paralysis.
- Open Access
Spiking Neural Network Decoders of Finger Forces from High-Density Intramuscular Microelectrode Arrays
Restoring naturalistic finger control in assistive technologies requires the continuous decoding of motor intent with high accuracy, efficiency, and robustness. Here, we present a spike-based decoding framework that integrates spiking neural networks (SNNs) with motor unit activity extracted from high-density intramuscular microelectrode arrays. We demonstrate simultaneous and proportional decoding of individual finger forces from motor unit spike trains during isometric contractions at 15% of maximum voluntary contraction using SNNs. We systematically evaluated alternative SNN decoder configurations and compared two possible input modalities: physiologically grounded motor unit spike trains and spike-encoded intramuscular EMG signals. Through this comparison, we quantified trade-offs between decoding accuracy, memory footprint, and robustness to input errors. The results showed that shallow SNNs can reliably decode finger-level motor intent with competitive accuracy and minimal latency, while operating with reduced memory requirements and without the need for external preprocessing buffers. This work provides a practical blueprint for integrating SNNs into finger-level force decoding systems, demonstrating how the choice of input representation can be strategically tailored to meet application-specific requirements for accuracy, robustness, and memory efficiency.
Open AccessSpiking Neural Network Decoders of Finger Forces from High-Density Intramuscular Microelectrode Arrays
Restoring naturalistic finger control in assistive technologies requires the continuous decoding of motor intent with high accuracy, efficiency, and robustness. Here, we present a spike-based decoding framework that integrates spiking neural networks (SNNs) with motor unit activity extracted from high-density intramuscular microelectrode arrays. We demonstrate simultaneous and proportional decoding of individual finger forces from motor unit spike trains during isometric contractions at 15% of maximum voluntary contraction using SNNs. We systematically evaluated alternative SNN decoder configurations and compared two possible input modalities: physiologically grounded motor unit spike trains and spike-encoded intramuscular EMG signals. Through this comparison, we quantified trade-offs between decoding accuracy, memory footprint, and robustness to input errors. The results showed that shallow SNNs can reliably decode finger-level motor intent with competitive accuracy and minimal latency, while operating with reduced memory requirements and without the need for external preprocessing buffers. This work provides a practical blueprint for integrating SNNs into finger-level force decoding systems, demonstrating how the choice of input representation can be strategically tailored to meet application-specific requirements for accuracy, robustness, and memory efficiency.
- Open Access
Multidimensional motoneuron control using intramuscular microelectrode arrays in tetraplegic spinal cord injury
Loss of hand function after spinal cord injury (SCI) significantly impairs independence and quality of life. Although residual muscle activity recorded on the skin can provide intuitive control signals, it is limited in SCI by low amplitude, and poor signal-to-noise ratio. Here, we show that these key limitations can be overcome by implanting 40-channel microelectrode arrays directly into forearm muscles. Using blind source separation methods, we detected the simultaneous activity of up to 73 spinal motoneurons during attempted movements of two tetraplegic individuals with SCI. Even in the presence of spasticity or complete paralysis, participants could modulate the activity of individual motoneurons to accurately control a computer cursor in one and two dimensions, play a video game, and even regain the ability to grasp using a motoneuron-controlled soft exoskeleton glove. Our findings demonstrate the feasibility of an implantable microelectrode array system that uses residual motoneuron activity for control in SCI, with the potential to improve the quality of life for individuals with paralysis.
Open AccessMultidimensional motoneuron control using intramuscular microelectrode arrays in tetraplegic spinal cord injury
Loss of hand function after spinal cord injury (SCI) significantly impairs independence and quality of life. Although residual muscle activity recorded on the skin can provide intuitive control signals, it is limited in SCI by low amplitude, and poor signal-to-noise ratio. Here, we show that these key limitations can be overcome by implanting 40-channel microelectrode arrays directly into forearm muscles. Using blind source separation methods, we detected the simultaneous activity of up to 73 spinal motoneurons during attempted movements of two tetraplegic individuals with SCI. Even in the presence of spasticity or complete paralysis, participants could modulate the activity of individual motoneurons to accurately control a computer cursor in one and two dimensions, play a video game, and even regain the ability to grasp using a motoneuron-controlled soft exoskeleton glove. Our findings demonstrate the feasibility of an implantable microelectrode array system that uses residual motoneuron activity for control in SCI, with the potential to improve the quality of life for individuals with paralysis.
- Top Cited PapersOpen Access
- 17 March 2025
- Published by Wiley in The Journal of Physiology
- Vol. 603 (8) , 2281-2300
Unlocking the full potential of high‐density surface EMG: novel non‐invasive high‐yield motor unit decomposition
The decomposition of high‐density surface electromyography (HD‐sEMG) signals into motor unit discharge patterns has become a powerful tool for investigating the neural control of movement, providing insights into motor neuron recruitment and discharge behaviour. However, current algorithms, while effective under certain conditions, face significant challenges in complex scenarios, as their accuracy and motor unit yield are highly dependent on anatomical differences among individuals. To address this issue, we recently introduced Swarm‐Contrastive Decomposition (SCD), which dynamically adjusts the contrast function based on the distribution of the data. Here, we demonstrate the ability of SCD in identifying low‐amplitude motor unit action potentials and effectively handling complex decomposition scenarios. We validated SCD using simulated and experimental HD‐sEMG recordings and compared it with current state‐of‐the‐art decomposition methods under varying conditions, including different excitation levels, noise intensities, force profiles, sexes and muscle groups. The proposed method consistently outperformed existing techniques in both the quantity of decoded motor units and the precision of their firing time identification. Across different simulated excitation levels, SCD detected, on average, 25.9 ±5.8 motor units vs. 13.9 ± 2.7 found by a state‐of‐the‐art baseline approach. Across noise levels, SCD detected 19.8 ± 13.5 motor units, compared to 11.9 ± 6.9 by the baseline method. In simulated conditions of high synchronisation levels, SCD detected approximately three times as many motor units compared to previous methods (31.2 ± 4.3 for SCD, 10.5 ± 1.7 for baseline), while also significantly improving accuracy. These advancements represent a step forward in non‐invasive EMG technology for studying motor unit activity in complex scenarios. image Key points: High‐density surface electromyography (HD‐sEMG) decomposition provides information on how the nervous system controls muscles, but current methods struggle in complex conditions. Swarm‐Contrastive Decomposition (SCD) is a new approach that dynamically adjusts how signals are separated, improving accuracy and increasing the sample of detected motor units. SCD successfully identifies more motor units, including those with low‐amplitude signals, and performs well even in challenging conditions such as high‐interference signals. In simulated ballistic contractions, SCD detected three times more motor units than previous methods while improving accuracy. These advancements could improve non‐invasive studies of muscle function in movement, fatigue and neurological disorders.
- 17 March 2025
- Published by Wiley in The Journal of Physiology
- Vol. 603 (8) , 2281-2300
Top Cited PapersOpen AccessUnlocking the full potential of high‐density surface EMG: novel non‐invasive high‐yield motor unit decomposition
The decomposition of high‐density surface electromyography (HD‐sEMG) signals into motor unit discharge patterns has become a powerful tool for investigating the neural control of movement, providing insights into motor neuron recruitment and discharge behaviour. However, current algorithms, while effective under certain conditions, face significant challenges in complex scenarios, as their accuracy and motor unit yield are highly dependent on anatomical differences among individuals. To address this issue, we recently introduced Swarm‐Contrastive Decomposition (SCD), which dynamically adjusts the contrast function based on the distribution of the data. Here, we demonstrate the ability of SCD in identifying low‐amplitude motor unit action potentials and effectively handling complex decomposition scenarios. We validated SCD using simulated and experimental HD‐sEMG recordings and compared it with current state‐of‐the‐art decomposition methods under varying conditions, including different excitation levels, noise intensities, force profiles, sexes and muscle groups. The proposed method consistently outperformed existing techniques in both the quantity of decoded motor units and the precision of their firing time identification. Across different simulated excitation levels, SCD detected, on average, 25.9 ±5.8 motor units vs. 13.9 ± 2.7 found by a state‐of‐the‐art baseline approach. Across noise levels, SCD detected 19.8 ± 13.5 motor units, compared to 11.9 ± 6.9 by the baseline method. In simulated conditions of high synchronisation levels, SCD detected approximately three times as many motor units compared to previous methods (31.2 ± 4.3 for SCD, 10.5 ± 1.7 for baseline), while also significantly improving accuracy. These advancements represent a step forward in non‐invasive EMG technology for studying motor unit activity in complex scenarios. image Key points: High‐density surface electromyography (HD‐sEMG) decomposition provides information on how the nervous system controls muscles, but current methods struggle in complex conditions. Swarm‐Contrastive Decomposition (SCD) is a new approach that dynamically adjusts how signals are separated, improving accuracy and increasing the sample of detected motor units. SCD successfully identifies more motor units, including those with low‐amplitude signals, and performs well even in challenging conditions such as high‐interference signals. In simulated ballistic contractions, SCD detected three times more motor units than previous methods while improving accuracy. These advancements could improve non‐invasive studies of muscle function in movement, fatigue and neurological disorders.
- Open Access
- 17 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 33 (15344320) , 620-629
Motor Unit Sampling From Intramuscular Micro-Electrode Array Recordings
Recordings of electrical activity from muscles allow us to identify the activity of pools of spinal motor neurons that send the neural drive for muscle activation. Decoding motor unit and motor neuron activity from muscle recordings can be performed by high-density (HD) electrode systems, both non-invasively (surface, HD-sEMG) and invasively (intramuscular, HD-iEMG). HD-sEMG recordings are obtained by grids placed on the skin surface while HD-iEMG signals can be acquired by micro-electrode arrays. While it has been shown that HD-iEMG allows the accurate decoding of a larger number of motor units when compared to HD-sEMG, the dependence of motor unit yield on the parameters of the micro-electrode arrays is still unexplored. Here, we used recently developed HD-iEMG electrodes to record from hundreds of recording sites within the muscle. This allowed us to investigate the impact of electrode number, inter-electrode distance, and the number of muscle insertions on the ability to sample motor units within the muscle. Specifically, we recorded both HD-sEMG and HD-iEMG from the Tibialis Anterior muscle of two healthy subjects at various contraction intensities (10%, 30%, and 70% of maximum voluntary contraction, MVC). For the first time, we present intramuscular recordings with more than 140 electrodes inside a single muscle, achieved through multiple implants of high-density micro-electrode arrays. Through systematic offline analyses of these recordings, we tested different electrode configurations to identify optimal setups for accurately capturing motor unit activity. The results revealed that the density of electrodes in the micro-electrode arrays is the most critical factor for maximising the number of identified motor units and ensuring very high accuracy. Comparisons between intramuscular and surface recordings also confirmed that HD-iEMG consistently captures larger and more stable numbers of motor units across subjects and contraction levels. These results underscore the potential of HD-iEMG as a powerful tool for both clinical and research settings, particularly when precise motor unit decomposition is crucial.
- 17 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 33 (15344320) , 620-629
Open AccessMotor Unit Sampling From Intramuscular Micro-Electrode Array Recordings
Recordings of electrical activity from muscles allow us to identify the activity of pools of spinal motor neurons that send the neural drive for muscle activation. Decoding motor unit and motor neuron activity from muscle recordings can be performed by high-density (HD) electrode systems, both non-invasively (surface, HD-sEMG) and invasively (intramuscular, HD-iEMG). HD-sEMG recordings are obtained by grids placed on the skin surface while HD-iEMG signals can be acquired by micro-electrode arrays. While it has been shown that HD-iEMG allows the accurate decoding of a larger number of motor units when compared to HD-sEMG, the dependence of motor unit yield on the parameters of the micro-electrode arrays is still unexplored. Here, we used recently developed HD-iEMG electrodes to record from hundreds of recording sites within the muscle. This allowed us to investigate the impact of electrode number, inter-electrode distance, and the number of muscle insertions on the ability to sample motor units within the muscle. Specifically, we recorded both HD-sEMG and HD-iEMG from the Tibialis Anterior muscle of two healthy subjects at various contraction intensities (10%, 30%, and 70% of maximum voluntary contraction, MVC). For the first time, we present intramuscular recordings with more than 140 electrodes inside a single muscle, achieved through multiple implants of high-density micro-electrode arrays. Through systematic offline analyses of these recordings, we tested different electrode configurations to identify optimal setups for accurately capturing motor unit activity. The results revealed that the density of electrodes in the micro-electrode arrays is the most critical factor for maximising the number of identified motor units and ensuring very high accuracy. Comparisons between intramuscular and surface recordings also confirmed that HD-iEMG consistently captures larger and more stable numbers of motor units across subjects and contraction levels. These results underscore the potential of HD-iEMG as a powerful tool for both clinical and research settings, particularly when precise motor unit decomposition is crucial.
- Open Access
- 20 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Motor unit sampling from intramuscular micro-electrode array recordings
Recordings of electrical activity from muscles allow us to identify the activity of pools of spinal motor neurons that send the neural drive for muscle activation. Decoding motor unit (MU) and motor neuron activity from muscle recordings can be performed by high-density (HD) electrode systems, both noninvasively (surface, HD-sEMG) and invasively (intramuscular, HD-iEMG). HD-sEMG recordings are obtained by grids placed on the skin surface while HD-iEMG signals can be acquired by microarrays of electrodes. While it has been shown that HD-iEMG allows the accurate decoding of a larger number of motor units when compared to HD-sEMG, the dependence of motor unit yield on the parameters of the micro-arrays is still unexplored. Here, we used recently developed HD-iEMG electrodes to record from hundreds of recording sites within the muscle. This allowed us to investigate the impact of electrode number, inter-electrode distance, and the number of muscle insertions on the ability to sample motor units within the muscle. Specifically, we recorded both HD-sEMG and HD-iEMG from the Tibialis Anterior muscle of two healthy subjects at various contraction intensities (10%, 30%, and 70% of maximum voluntary contraction, MVC). For the first time, we present intramuscular recordings with more than 140 electrodes inside a single muscle, achieved through multiple implants of highdensity micro-arrays. Through systematic offline analyses of these recordings, we tested different electrode configurations to identify optimal setups for accurately capturing motor unit activity. The results revealed that the density of electrodes in the micro-arrays is the most critical factor for maximising the number of identified motor units and ensuring very high accuracy. Comparisons between intramuscular and surface recordings also confirmed that HD-iEMG consistently captures larger and more stable numbers of motor units across subjects and contraction levels. These results underscore the potential of HD-iEMG as a powerful tool for both clinical and research settings, particularly when precise motor unit decomposition is crucial.
- 20 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessMotor unit sampling from intramuscular micro-electrode array recordings
Recordings of electrical activity from muscles allow us to identify the activity of pools of spinal motor neurons that send the neural drive for muscle activation. Decoding motor unit (MU) and motor neuron activity from muscle recordings can be performed by high-density (HD) electrode systems, both noninvasively (surface, HD-sEMG) and invasively (intramuscular, HD-iEMG). HD-sEMG recordings are obtained by grids placed on the skin surface while HD-iEMG signals can be acquired by microarrays of electrodes. While it has been shown that HD-iEMG allows the accurate decoding of a larger number of motor units when compared to HD-sEMG, the dependence of motor unit yield on the parameters of the micro-arrays is still unexplored. Here, we used recently developed HD-iEMG electrodes to record from hundreds of recording sites within the muscle. This allowed us to investigate the impact of electrode number, inter-electrode distance, and the number of muscle insertions on the ability to sample motor units within the muscle. Specifically, we recorded both HD-sEMG and HD-iEMG from the Tibialis Anterior muscle of two healthy subjects at various contraction intensities (10%, 30%, and 70% of maximum voluntary contraction, MVC). For the first time, we present intramuscular recordings with more than 140 electrodes inside a single muscle, achieved through multiple implants of highdensity micro-arrays. Through systematic offline analyses of these recordings, we tested different electrode configurations to identify optimal setups for accurately capturing motor unit activity. The results revealed that the density of electrodes in the micro-arrays is the most critical factor for maximising the number of identified motor units and ensuring very high accuracy. Comparisons between intramuscular and surface recordings also confirmed that HD-iEMG consistently captures larger and more stable numbers of motor units across subjects and contraction levels. These results underscore the potential of HD-iEMG as a powerful tool for both clinical and research settings, particularly when precise motor unit decomposition is crucial.
- 21 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 72 (1) , 227-237
A Particle Swarm Optimized Independence Estimator for Blind Source Separation of Neurophysiological Time Series
The decomposition of neurophysiological recordings into their constituent neural sources is of major importance to a diverse range of neuroscientific fields and neuroengineering applications. The advent of high density electrode probes and arrays has driven a major need for novel semi-automated and automated blind source separation methodologies that take advantage of the increased spatial resolution and coverage these new devices offer. Independent component analysis (ICA) offers a principled theoretical framework for such algorithms, but implementation inefficiencies often drive poor performance in practice, particularly for sparse sources. Here we observe that the use of a single non-linear optimization function to identify spiking sources with ICA often has a detrimental effect that precludes the recovery and correct separation of all spiking sources in the signal. We go on to propose a projection-pursuit ICA algorithm designed specifically for spiking sources, which uses a particle swarm methodology to adaptively traverse a polynomial family of non-linearities approximating the asymmetric cumulants of the sources. We robustly prove state-of-the-art decomposition performance on recordings from high density intramuscular probes and demonstrate how the particle swarm quickly finds optimal contrast non-linearities across a range of neurophysiological datasets.
- 21 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 72 (1) , 227-237
A Particle Swarm Optimized Independence Estimator for Blind Source Separation of Neurophysiological Time Series
The decomposition of neurophysiological recordings into their constituent neural sources is of major importance to a diverse range of neuroscientific fields and neuroengineering applications. The advent of high density electrode probes and arrays has driven a major need for novel semi-automated and automated blind source separation methodologies that take advantage of the increased spatial resolution and coverage these new devices offer. Independent component analysis (ICA) offers a principled theoretical framework for such algorithms, but implementation inefficiencies often drive poor performance in practice, particularly for sparse sources. Here we observe that the use of a single non-linear optimization function to identify spiking sources with ICA often has a detrimental effect that precludes the recovery and correct separation of all spiking sources in the signal. We go on to propose a projection-pursuit ICA algorithm designed specifically for spiking sources, which uses a particle swarm methodology to adaptively traverse a polynomial family of non-linearities approximating the asymmetric cumulants of the sources. We robustly prove state-of-the-art decomposition performance on recordings from high density intramuscular probes and demonstrate how the particle swarm quickly finds optimal contrast non-linearities across a range of neurophysiological datasets.
- Open Access
Design, Fabrication and Evaluation of a Stretchable High-Density Electromyography Array
The adoption of high-density electrode systems for human–machine interfaces in real-life applications has been impeded by practical and technical challenges, including noise interference, motion artefacts and the lack of compact electrode interfaces. To overcome some of these challenges, we introduce a wearable and stretchable electromyography (EMG) array, and present its design, fabrication methodology, characterisation, and comprehensive evaluation. Our proposed solution comprises dry-electrodes on flexible printed circuit board (PCB) substrates, eliminating the need for time-consuming skin preparation. The proposed fabrication method allows the manufacturing of stretchable sleeves, with consistent and standardised coverage across subjects. We thoroughly tested our developed prototype, evaluating its potential for application in both research and real-world environments. The results of our study showed that the developed stretchable array matches or outperforms traditional EMG grids and holds promise in furthering the real-world translation of high-density EMG for human–machine interfaces.
Open AccessDesign, Fabrication and Evaluation of a Stretchable High-Density Electromyography Array
The adoption of high-density electrode systems for human–machine interfaces in real-life applications has been impeded by practical and technical challenges, including noise interference, motion artefacts and the lack of compact electrode interfaces. To overcome some of these challenges, we introduce a wearable and stretchable electromyography (EMG) array, and present its design, fabrication methodology, characterisation, and comprehensive evaluation. Our proposed solution comprises dry-electrodes on flexible printed circuit board (PCB) substrates, eliminating the need for time-consuming skin preparation. The proposed fabrication method allows the manufacturing of stretchable sleeves, with consistent and standardised coverage across subjects. We thoroughly tested our developed prototype, evaluating its potential for application in both research and real-world environments. The results of our study showed that the developed stretchable array matches or outperforms traditional EMG grids and holds promise in furthering the real-world translation of high-density EMG for human–machine interfaces.
- Open Access
- 13 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
A Particle Swarm Optimised Independence Estimator for Blind Source Separation of Neurophysiological Time Series
The decomposition of neurophysiological recordings into their constituent neural sources is of major importance to a diverse range of neuroscientific fields and neuroengineering applications. The advent of high density electrode probes and arrays has driven a major need for novel semi-automated and automated blind source separation methodologies that take advantage of the increased spatial resolution and coverage these new devices offer. Independent component analysis (ICA) offers a principled theoretical framework for such algorithms, but implementation inefficiencies often drive poor performance in practice, particularly for sparse sources. Here we observe that the use of a single non-linear optimization function to identify spiking sources with ICA often has a detrimental effect that precludes the recovery and correct separation of all spiking sources in the signal. We go on to propose a projection-pursuit ICA algorithm designed specifically for spiking sources, which uses a particle swarm methodology to adaptively traverse a polynomial family of non-linearities approximating the asymmetric cumulants of the sources. We robustly prove state-of-the-art decomposition performance on recordings from high density intramuscular probes and demonstrate how the particle swarm quickly finds optimal contrast non-linearities across a range of neurophysiological datasets.
- 13 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessA Particle Swarm Optimised Independence Estimator for Blind Source Separation of Neurophysiological Time Series
The decomposition of neurophysiological recordings into their constituent neural sources is of major importance to a diverse range of neuroscientific fields and neuroengineering applications. The advent of high density electrode probes and arrays has driven a major need for novel semi-automated and automated blind source separation methodologies that take advantage of the increased spatial resolution and coverage these new devices offer. Independent component analysis (ICA) offers a principled theoretical framework for such algorithms, but implementation inefficiencies often drive poor performance in practice, particularly for sparse sources. Here we observe that the use of a single non-linear optimization function to identify spiking sources with ICA often has a detrimental effect that precludes the recovery and correct separation of all spiking sources in the signal. We go on to propose a projection-pursuit ICA algorithm designed specifically for spiking sources, which uses a particle swarm methodology to adaptively traverse a polynomial family of non-linearities approximating the asymmetric cumulants of the sources. We robustly prove state-of-the-art decomposition performance on recordings from high density intramuscular probes and demonstrate how the particle swarm quickly finds optimal contrast non-linearities across a range of neurophysiological datasets.