Articles publiés dans des revues avec comité de lecture

URI permanent de cette collection

Parcourir par

Dépôts récents

Voici les éléments 1 - 5 sur 5198
  • Publication
    Accès libre
    Interpreting deep learning features for myoelectric control : a comparison with handcrafted features
    (Frontiers Media, 2020-03-03) Côté Allard, Ulysse; Campbell, Evan; Phinyomark, Angkoon; Laviolette, François; Gosselin, Benoit; Scheme, Erik
    Existing research on myoelectric control systems primarily focuses on extracting discriminative characteristics of the electromyographic (EMG) signal by designing handcrafted features. Recently, however, deep learning techniques have been applied to the challenging task of EMG-based gesture recognition. The adoption of these techniques slowly shifts the focus from feature engineering to feature learning. Nevertheless, the black-box nature of deep learning makes it hard to understand the type of information learned by the network and how it relates to handcrafted features. Additionally, due to the high variability in EMG recordings between participants, deep features tend to generalize poorly across subjects using standard training methods. Consequently, this work introduces a new multi-domain learning algorithm, named ADANN (Adaptive Domain Adversarial Neural Network), which significantly enhances (p = 0.00004) inter-subject classification accuracy by an average of 19.40% compared to standard training. Using ADANN-generated features, this work provides the first topological data analysis of EMG-based gesture recognition for the characterization of the information encoded within a deep network, using handcrafted features as landmarks. This analysis reveals that handcrafted features and the learned features (in the earlier layers) both try to discriminate between all gestures, but do not encode the same information to do so. In the later layers, the learned features are inclined to instead adopt a one-vs.-all strategy for a given class. Furthermore, by using convolutional network visualization techniques, it is revealed that learned features actually tend to ignore the most activated channel during contraction, which is in stark contrast with the prevalence of handcrafted features designed to capture amplitude information. Overall, this work paves the way for hybrid feature sets by providing a clear guideline of complementary information encoded within learned and handcrafted features.
  • Publication
    Accès libre
    A wireless electro-optic platform for multimodal electrophysiology and optogenetics in freely moving rodents
    (Frontiers Media S.A., 2021-08-16) Bilodeau, Guillaume; Gagnon-Turcotte, Gabriel; L. Gagnon, Léonard; Keramidis, Iason; Timofeev, Igor; De Koninck, Yves; Ethier, Christian; Gosselin, Benoit
    This paper presents the design and the utilization of a wireless electro-optic platform to perform simultaneous multimodal electrophysiological recordings and optogenetic stimulation in freely moving rodents. The developed system can capture neural action potentials (AP), local field potentials (LFP) and electromyography (EMG) signals with up to 32 channels in parallel while providing four optical stimulation channels. The platform is using commercial off-the-shelf components (COTS) and a low-power digital field-programmable gate array (FPGA), to perform digital signal processing to digitally separate in real time the AP, LFP and EMG while performing signal detection and compression for mitigating wireless bandwidth and power consumption limitations. The different signal modalities collected on the 32 channels are time-multiplexed into a single data stream to decrease power consumption and optimize resource utilization. The data reduction strategy is based on signal processing and real-time data compression. Digital filtering, signal detection, and wavelet data compression are used inside the platform to separate the different electrophysiological signal modalities, namely the local field potentials (1–500 Hz), EMG (30–500 Hz), and the action potentials (300–5,000 Hz) and perform data reduction before transmitting the data. The platform achieves a measured data reduction ratio of 7.77 (for a firing rate of 50 AP/second) and weights 4.7 g with a 100-mAh battery, an on/off switch and a protective plastic enclosure. To validate the performance of the platform, we measured distinct electrophysiology signals and performed optogenetics stimulation in vivo in freely moving rondents. We recorded AP and LFP signals with the platform using a 16-microelectrode array implanted in the primary motor cortex of a Long Evans rat, both in anesthetized and freely moving conditions. EMG responses to optogenetic Channelrhodopsin-2 induced activation of motor cortex via optical fiber were also recorded in freely moving rodents.
  • Publication
    Restreint
    A wireless electro-optic headstage with a 0.13-μm CMOS customintegrated DWT neural signal decoder for closed-loop optogenetics
    (IEEE, 2019-07-23) Gagnon-Turcotte, Gabriel; Keramidis, Iason; Ethier, Christian; De Koninck, Yves; Gosselin, Benoit
    We present awireless electro-optic headstage that uses a 0.13-μm CMOS custom integrated circuit (IC) implementing a digital neural decoder (ND-IC) for enabling real-time closed-loop (CL) optogenetics. The ND-IC processes the neural activity data using three digital cores: 1) the detector core detects and extracts the action potential (AP) of individual neurons by using an adaptive threshold; 2) the data compression core compresses the detected AP by using an efficient Symmlet-2 discrete wavelet transform (DWT) processor for decreasing the amount of data to be transmitted by the low-power wireless link; and 3) the classification core sorts the compressed AP into separated clusters on the fly according to their wave shapes. The ND-IC encompasses several innovations: 1) the compression core decreases the complexity from O(n2) to O(n· log(n)) compared to the previous solutions, while using two times less memory, thanks to the use of a new coefficient sorting tree; and 2) the AP classification core reuses both the compressed DWT coefficients to perform implicit dimensionality reduction, which allows for performing intensive signal processing on-chip, while increasing power and hardware efficiency. This core also reuses the signal standard deviation already computed by theAPdetector core as threshold for performing automatic AP sorting. The headstage also introduces innovations by enabling a new wireless CL scheme between the neural data acquisition module and the optical stimulator. Our CL scheme uses the AP sorting and timing information produced by the ND-IC for detecting complex firing patternswithin the brain. The headstage is also smaller (1.13 cm3), lighter (3.0 g with a 40mAhbattery) and less invasive than the previous solutions, while providing a measured autonomy of 2h40, with the ND-IC. The whole system and the ND-IC are first validated in vivo in the LD thalamus of a Long-Evans rat, and then in freely-moving CL experiments involving a mouse virally expressing ChR2-mCherry in inhibitory neurons of the prelimbic cortex, and the results show that our system works well within an in vivo experimental setting with a freely moving mouse.
  • Publication
    Restreint
    A wireless headstage for combined optogenetics and multichannel electrophysiological recording
    (IEEE, 2016-06-20) Gagnon-Turcotte, Gabriel; LeChasseur, Yoan; Bories, Cyril; Messaddeq, Younès; De Koninck, Yves; Gosselin, Benoit
    This paper presents a wireless headstage with realtime spike detection and data compression for combined optogenetics and multichannel electrophysiological recording. The proposed headstage, which is intended to perform both optical stimulation and electrophysiological recordings simultaneously in freely moving transgenic rodents, is entirely built with commercial off-the-shelf components, and includes 32 recording channels and 32 optical stimulation channels. It can detect, compress and transmit full action potential waveforms over 32 channels in parallel and in real time using an embedded digital signal processor based on a low-power field programmable gate array and a Microblaze microprocessor softcore. Such a processor implements a complete digital spike detector featuring a novel adaptive threshold based on a Sigma-delta control loop, and a wavelet data compression module using a new dynamic coefficient re-quantization technique achieving large compression ratios with higher signal quality. Simultaneous optical stimulation and recording have been performed in-vivo using an optrode featuring 8 microelectrodes and 1 implantable fiber coupled to a 465-nm LED, in the somatosensory cortex and the Hippocampus of a transgenic mouse expressing ChannelRhodospin (Thy1::ChR2-YFP line 4) under anesthetized conditions. Experimental results show that the proposed headstage can trigger neuron activity while collecting, detecting and compressing single cell microvolt amplitude activity from multiple channels in parallel while achieving overall compression ratios above 500. This is the first reported high-channel count wireless optogenetic device providing simultaneous optical stimulation and recording. Measured characteristics show that the proposed headstage can achieve up to 100% of true positive detection rate for signal-to-noise ratio (SNR) down to 15 dB, while achieving up to 97.28% at SNR as low as 5 dB. The implemented prototype features a lifespan of up to 105 minutes, and uses a lightweight (2.8 g) and compact (17 × 18 × 10 mm3) rigid-flex printed circuit board.
  • Publication
    Accès libre
    A low-cost, wireless, 3-D-printed custom armband for sEMG hand gesture recognition
    (MDPI, 2019-06-24) Côté Allard, Ulysse; Gagnon-Turcotte, Gabriel; Laviolette, François; Gosselin, Benoit
    Wearable technology can be employed to elevate the abilities of humans to perform demanding and complex tasks more efficiently. Armbands capable of surface electromyography (sEMG) are attractive and noninvasive devices from which human intent can be derived by leveraging machine learning. However, the sEMG acquisition systems currently available tend to be prohibitively costly for personal use or sacrifice wearability or signal quality to be more affordable. This work introduces the 3DC Armband designed by the Biomedical Microsystems Laboratory in Laval University; a wireless, 10-channel, 1000 sps, dry-electrode, low-cost ( 150 USD) myoelectric armband that also includes a 9-axis inertial measurement unit. The proposed system is compared with the Myo Armband by Thalmic Labs, one of the most popular sEMG acquisition systems. The comparison is made by employing a new offline dataset featuring 22 able-bodied participants performing eleven hand/wrist gestures while wearing the two armbands simultaneously. The 3DC Armband systematically and significantly (p < 0.05) outperforms the Myo Armband, with three different classifiers employing three different input modalities when using ten seconds or more of training data per gesture. This new dataset, alongside the source code, Altium project and 3-D models are made readily available for download within a Github repository.