Decoding the brain: From neural representations to mechanistic models

Cell, Год журнала: 2024, Номер 187(21), С. 5814 - 5832

Опубликована: Окт. 1, 2024

Язык: Английский

---

Multiscale fusion enhanced spiking neural network for invasive BCI neural signal decoding

Frontiers in Neuroscience, Год журнала: 2025, Номер 19

Опубликована: Фев. 21, 2025

Brain-computer interfaces (BCIs) are an advanced fusion of neuroscience and artificial intelligence, requiring stable long-term decoding neural signals. Spiking Neural Networks (SNNs), with their neuronal dynamics spike-based signal processing, inherently well-suited for this task. This paper presents a novel approach utilizing Multiscale Fusion enhanced Network (MFSNN). The MFSNN emulates the parallel processing multiscale feature seen in human visual perception to enable real-time, efficient, energy-conserving decoding. Initially, employs temporal convolutional networks channel attention mechanisms extract spatiotemporal features from raw data. It then enhances performance by integrating these through skip connections. Additionally, improves generalizability robustness cross-day mini-batch supervised generalization learning. In two benchmark invasive BCI paradigms, including single-hand grasp-and-touch center-and-out reach tasks, surpasses traditional network methods, such as MLP GRU, both accuracy computational efficiency. Moreover, MFSNN's framework is implementation on neuromorphic chips, offering energy-efficient solution online

Язык: Английский

---

Modeling conditional distributions of neural and behavioral data with masked variational autoencoders

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Апрель 25, 2024

Extracting the relationship between high-dimensional recordings of neural activity and complex behavior is a ubiquitous problem in systems neuroscience. Toward this goal, encoding decoding models attempt to infer conditional distribution given vice versa, while dimensionality reduction techniques aim extract interpretable low-dimensional representations. Variational autoencoders (VAEs) are flexible deep-learning commonly used embeddings or behavioral data. However, it challenging for VAEs accurately model arbitrary distributions, such as those encountered decoding, even more so simultaneously. Here, we present VAE-based approach calculating distributions. We validate our on task with known ground truth demonstrate applicability time series by retrieving distributions over masked body parts walking flies. Finally, probabilistically decode motor trajectories from population monkey reach query same VAE behavior. Our provides unifying perspective joint learning data, which will allow scaling common analyses neuroscience today's multi-modal datasets.

Язык: Английский

---

From monkeys to humans: observation-based EMG brain–computer interface decoders for humans with paralysis

Journal of Neural Engineering, Год журнала: 2023, Номер 20(5), С. 056040 - 056040

Опубликована: Окт. 1, 2023

. Intracortical brain-computer interfaces (iBCIs) aim to enable individuals with paralysis control the movement of virtual limbs and robotic arms. Because patients' prevents training a direct neural activity limb decoder, most iBCIs rely on 'observation-based' decoding in which patient watches moving cursor while mentally envisioning making movement. However, this reliance observed target motion for decoder development precludes its application prediction unobservable motor output like muscle activity. Here, we ask whether recordings from surrogate individual performing same as iBCI can be used an decoder.

Язык: Английский

---

3D-aware neural network for analyzing neuron morphology

Опубликована: Апрель 12, 2024

Язык: Английский

---

Few-shot Algorithms for Consistent Neural Decoding (FALCON) Benchmark

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Сен. 16, 2024

Abstract Intracortical brain-computer interfaces (iBCIs) can restore movement and communication abilities to individuals with paralysis by decoding their intended behavior from neural activity recorded an implanted device. While this yields high-performance over short timescales, data are often nonstationary, which lead decoder failure if not accounted for. To maintain performance, users must frequently recalibrate decoders, requires the arduous collection of new behavioral data. Aiming reduce burden, several approaches have been developed that either limit recalibration requirements (few-shot approaches) or eliminate explicit entirely (zero-shot approaches). However, progress is limited a lack standardized datasets comparison metrics, causing methods be compared in ad hoc manner. Here we introduce FALCON benchmark suite (Few-shot Algorithms for COnsistent Neural decoding) standardize evaluation iBCI robustness. curates five span tasks focus on behaviors interest modern-day iBCIs. Each dataset includes calibration data, optional few-shot private We implement flexible platform only user-submitted code return predictions unseen also seed applying baseline spanning classes possible approaches. aims provide rigorous selection criteria robust easing translation real-world devices. https://snel-repo.github.io/falcon/

Язык: Английский

---

Unsupervised, piecewise linear decoding enables an accurate prediction of muscle activity in a multi-task brain computer interface

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Сен. 13, 2024

Creating an intracortical brain-computer interface (iBCI) capable of seamless transitions between tasks and contexts would greatly enhance user experience. However, the nonlinearity in neural activity presents challenges to computing a global iBCI decoder. We aimed develop method that differs from globally optimized decoder address this issue.

Язык: Английский

---