About

Lee Miller is a Professor of Physiology, a Professor of Physical Medicine & Rehabilitation, and a Professor of Biomedical Engineering (by courtesy) at Northwestern University. His research focuses on understanding the signals produced by neurons in the central nervous system, particularly how the brain encodes movement commands. His work aims to decipher the nature of these signals, the mechanisms by which they are generated, and how they can be applied therapeutically to human patients. Much of his research involves recording neural activity directly from the brains of animals during behavior to study the signals produced by individual neurons within complex neural circuits. Miller's research has demonstrated that neurons in the motor cortex encode activity of small muscle groups that act synergistically to control movement, with different neurons controlling different muscle groups to produce coordinated actions. He has also studied brainstem motor areas such as the red nucleus and the superior colliculus, which have specialized functions related to hand and shoulder movements. His laboratory employs advanced electrode arrays to record from numerous neurons simultaneously, developing computational tools to analyze their functional connectivity. His work has contributed to the emerging field of Brain Machine Interface (BMI), where recordings from the brain are used to control external devices or restore muscle function, including efforts to develop BMIs that bypass injured spinal cords to activate muscles directly. His research combines technological innovation with fundamental neuroscience to advance understanding of motor control and develop potential therapeutic applications.

Research topics

• Artificial Intelligence
• Computer Science
• Neuroscience
• Biology
• Machine Learning
• Anatomy
• Psychology
• Simulation
• Physics
• Telecommunications

Selected publications

• Stabilizing brain-computer interfaces through alignment of latent dynamics

  Nature Communications · 2025-05-19 · 25 citations

  articleOpen access

  Intracortical brain-computer interfaces (iBCIs) restore motor function to people with paralysis by translating brain activity into control signals for external devices. In current iBCIs, instabilities at the neural interface result in a degradation of decoding performance, which necessitates frequent supervised recalibration using new labeled data. One potential solution is to use the latent manifold structure that underlies neural population activity to facilitate a stable mapping between brain activity and behavior. Recent efforts using unsupervised approaches have improved iBCI stability using this principle; however, existing methods treat each time step as an independent sample and do not account for latent dynamics. Dynamics have been used to enable high-performance prediction of movement intention, and may also help improve stabilization. Here, we present a platform for Nonlinear Manifold Alignment with Dynamics (NoMAD), which stabilizes decoding using recurrent neural network models of dynamics. NoMAD uses unsupervised distribution alignment to update the mapping of nonstationary neural data to a consistent set of neural dynamics, thereby providing stable input to the decoder. In applications to data from monkey motor cortex collected during motor tasks, NoMAD enables accurate behavioral decoding with unparalleled stability over weeks- to months-long timescales without any supervised recalibration. Current intracortical brain-computer interfaces are subject to recording interface instabilities that degrade decoding performance. Here, the authors present a platform for Nonlinear Manifold Alignment with Dynamics (NoMAD), which stabilizes decoding using models of dynamics for at least 3 months.

  PublisherOA PDFDOI

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

  Journal of Neural Engineering · 2025-01-17 · 2 citations

  articleOpen accessSenior author

  Abstract Objective. 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 to develop a method that differs from a globally optimized decoder to address this issue. Approach. We devised an unsupervised approach that relies on the structure of a low-dimensional neural manifold to implement a piecewise linear decoder. We created a distinctive dataset in which monkeys performed a diverse set of tasks, some trained, others innate, while we recorded neural signals from the motor cortex (M1) and electromyographs (EMGs) from upper limb muscles. We used both linear and nonlinear dimensionality reduction techniques to discover neural manifolds and applied unsupervised algorithms to identify clusters within those spaces. Finally, we fit a linear decoder of EMG for each cluster. A specific decoder was activated corresponding to the cluster each new neural data point belonged to. Main results. We found clusters in the neural manifolds corresponding with the different tasks or task sub-phases. The performance of piecewise decoding improved as the number of clusters increased and plateaued gradually. With only two clusters it already outperformed a global linear decoder, and unexpectedly, it outperformed even a global recurrent neural network decoder with 10–12 clusters. Significance. This study introduced a computationally lightweight solution for creating iBCI decoders that can function effectively across a broad range of tasks. EMG decoding is particularly challenging, as muscle activity is used, under varying contexts, to control interaction forces and limb stiffness, as well as motion. The results suggest that a piecewise linear decoder can provide a good approximation to the nonlinearity between neural activity and motor outputs, a result of our increased understanding of the structure of neural manifolds in motor cortex.

  PublisherDOI

• Tactile edges and motion via patterned microstimulation of the human somatosensory cortex

  Science · 2025-01-16 · 45 citations

  articleOpen access

  Intracortical microstimulation (ICMS) of somatosensory cortex evokes tactile sensations whose properties can be systematically manipulated by varying stimulation parameters. However, ICMS currently provides an imperfect sense of touch, limiting manual dexterity and tactile experience. Leveraging our understanding of how tactile features are encoded in the primary somatosensory cortex (S1), we sought to inform individuals with paralysis about local geometry and apparent motion of objects on their skin. We simultaneously delivered ICMS through electrodes with spatially patterned projected fields (PFs), evoking sensations of edges. We then created complex PFs that encode arbitrary tactile shapes and skin indentation patterns. By delivering spatiotemporally patterned ICMS, we evoked sensation of motion across the skin, the speed and direction of which could be controlled. Thus, we improved individuals' tactile experience and use of brain-controlled bionic hands.

  PublisherOA PDFDOI

• A Generalist Intracortical Motor Decoder

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-02-06 · 4 citations

  preprintOpen access

  Mapping the relationship between neural activity and motor behavior is a central aim of sensorimotor neuroscience and neurotechnology. While most progress to this end has relied on restricting complexity, the advent of foundation models instead proposes integrating a breadth of data as an alternate avenue for broadly advancing downstream modeling. We quantify this premise for motor decoding from intracortical microelectrode data, pretraining an autoregressive Transformer on 2000 hours of neural population spiking activity paired with diverse motor covariates from over 30 monkeys and humans. The resulting model is broadly useful, benefiting decoding on 8 downstream decoding tasks and generalizing to a variety of neural distribution shifts. However, we also highlight that scaling autoregressive Transformers seems unlikely to resolve limitations stemming from sensor variability and output stereotypy in neural datasets. Code: https://github.com/joel99/ndt3.

  PublisherOA PDFDOI

• Simulating the people's voice: Leveraging algorithmic fidelity to assess ChatGPT's performance in modeling public opinion on Chinese government policies

  Information Processing & Management · 2025-12-20 · 1 citations

  articleSenior authorCorresponding
  PublisherDOI

• Powering Small Satellites with Advanced NiH2 Dependent Pressure Vessel (DPV) Batteries

  Digital Commons - USU (Utah State University) · 2025-08-25

  articleOpen accessSenior author

  The Dependent Pressure Vessel (DPV) nickel-hydrogen (NiH2) design is being developed by Eagle-Picher Industries, Inc. (EPI), as a spacecraft battery for both large and small, military and commercial satellites. The DPV cell design offers high specific energy, energy density and reduced cost, while retaining the established IPV technology flight heritage and database. This advanced design also offers a more efficient mechanical, electrical and thermal cell and battery configuration and a reduced parts count. The geometry of the DPV cell promotes compact, minimum volume packaging and battery weight efficiency. The DPV battery design offers significant cost and weight savings potential while providing minimal design risks. In this presentation, we will discuss design features and present test data from existing development cells and address issues relevant to design and production of a DPV battery suitable for a small satellite application which would retain the energy increases and weight and cost reductions proposed. With the DPV, EPI has combined the unique features and significant advantages of NiH2 electrochemistry with the simplicity and extensive design heritage of the NiCd battery system.

  PublisherOA PDFDOI

• Amorphous silicon carbide probe mechanics for insertion in the cerebral cortex of rats, pigs, and macaques

  Journal of Neural Engineering · 2025-05-21 · 4 citations

  articleOpen access

  Abstract Objective. Intracortical microelectrode arrays (MEAs) are implantable devices used for neural recording and stimulation. However, their long-term performance is often compromised due, at least in part, to glial scar formation, initiated by microglial migration and astrocyte activation following implantation. To address this issue, ultra-thin MEAs (UMEAs) have been proposed as an alternative due to their reduced cross-sectional area (CSA) and enhanced flexibility that minimizes the mechanical mismatch between brain tissue and the electrode. These properties are expected to mitigate the persistent foreign-body response associated with micromotion. However, unaided implantation of UMEAs can be challenging, as their high flexibility may increase the likelihood of buckling and curtail the precise penetration into the brain. Approach. We investigated flexible amorphous silicon carbide (a-SiC) probe designs with varying CSAs and lengths to identify geometries that enable unsupported implantation into the cerebral cortices of rat, pig, and macaques. The critical buckling force of the a-SiC probes was experimentally determined as a function of geometry and formally described via finite element modeling, which predicted buckling behavior. Additionally, the penetration behavior of a-SiC probes was evaluated by measuring force–displacement responses during insertion into cortical tissues across species. Main results. Our findings demonstrated that the penetration force and cortical dimpling depth were not significantly influenced by the range of probe geometries tested. However, we observed that penetration force and dimpling depth were significantly lower in rat cortices than in larger species. Importantly, probe geometries with a higher ratio (⩾2) of critical buckling force to penetration force exhibited a 100% success rate for unaided insertion. Significance . This study provides a framework for designing and evaluating UMEA geometries to optimize unsupported implantation in both small and large animal brains.

  PublisherDOI

• Intracortical microstimulation in humans: a decade of safety and efficacy

  medRxiv · 2025-08-13 · 3 citations

  preprintOpen access

  Background: Intracortical microstimulation (ICMS) of somatosensory cortex can restore a sense of touch to people with spinal cord injury (SCI). In this early-feasibility clinical trial, we evaluate the safety, efficacy, and longevity of ICMS as there is a paucity of such long-term studies in humans. This information is crucial to the development of clinical neuromodulation devices, particularly for restoring touch, hearing, and vision. Methods: ICMS was delivered to five participants with SCI who were each implanted with two Blackrock NeuroPort microelectrode arrays in the hand representation of Brodmann's Area 1. Across implant durations spanning two to ten years, we measured single-electrode detection thresholds, projected fields, quality reports, and electrode health characteristics. ICMS-related adverse events were documented throughout. Results: Over 168 million ICMS pulses were delivered across a combined implant duration of 24 years without any serious adverse events or direct negative effects on electrode health. ICMS consistently evoked sensations localized to the hand. Rarely, sensations persisted for brief periods after stimulation offset (3-25 events across participants). ICMS detection thresholds increased slowly over time (~3.5 μA/year), but 62±15% of the electrodes still reliably evoke tactile sensations (~25% decrease in functional electrodes), including 55% of the electrodes after 10 years in one participant. The quality and projected field coverage of ICMS-evoked sensations were both consistent. Conclusions: Delivering ICMS to somatosensory cortex was safe and reliable, consistently evoking informative somatosensory percepts as long as 10 years, demonstrating the clinical promise of ICMS for sensory restoration. (Funded by NIH; ClinicalTrials.gov: NCT01894802).

  PublisherOA PDFDOI

• A neural implementation model of feedback-based motor learning

  Nature Communications · 2025-02-20 · 27 citations

  articleOpen access

  Animals use feedback to rapidly correct ongoing movements in the presence of a perturbation. Repeated exposure to a predictable perturbation leads to behavioural adaptation that compensates for its effects. Here, we tested the hypothesis that all the processes necessary for motor adaptation may emerge as properties of a controller that adaptively updates its policy. We trained a recurrent neural network to control its own output through an error-based feedback signal, which allowed it to rapidly counteract external perturbations. Implementing a biologically plausible plasticity rule based on this same feedback signal enabled the network to learn to compensate for persistent perturbations through a trial-by-trial process. The network activity changes during learning matched those from populations of neurons from monkey primary motor cortex - known to mediate both movement correction and motor adaptation - during the same task. Furthermore, our model natively reproduced several key aspects of behavioural studies in humans and monkeys. Thus, key features of trial-by-trial motor adaptation can arise from the internal properties of a recurrent neural circuit that adaptively controls its output based on ongoing feedback.

  PublisherOA PDFDOI

• Less is more: selection from a small set of options improves BCI velocity control

  Journal of Neural Engineering · 2025-03-05 · 1 citations

  articleOpen access

  Abstract Objective. Decoding algorithms used in invasive brain–computer interfaces (iBCIs) typically convert neural activity into continuously varying velocity commands. We hypothesized that putting constraints on which decoded velocity commands are permissible could improve user performance. To test this hypothesis, we designed the discrete direction selection (DDS) decoder, which uses neural activity to select among a small menu of preset cursor velocities. Approach . We tested DDS in a closed-loop cursor control task against many common continuous velocity decoders in both a human-operated real-time iBCI simulator (the jaBCI) and in a monkey using an iBCI. In the jaBCI, we compared performance across four visits by each of 48 naïve, able-bodied human subjects using either DDS, direct regression with assist (an affine map from neural activity to cursor velocity, DR-A), ReFIT, or the velocity Kalman Filter (vKF). In a follow up study to verify the jaBCI results, we compared a monkey’s performance using an iBCI with either DDS or the Wiener filter decoder (a direct regression decoder that includes time history, WF). Main Result . In the jaBCI, DDS substantially outperformed all other decoders with 93% mean targets hit per visit compared to DR-A, ReFIT, and vKF with 56%, 39%, and 26% mean targets hit, respectively. With the iBCI, the monkey achieved a 61% success rate with DDS and a 37% success rate with WF. Significance . Discretizing the decoded velocity with DDS effectively traded high resolution velocity commands for less tortuous and lower noise trajectories, highlighting the potential benefits of discretization in simplifying online BCI control.

  PublisherDOI

Recent grants

• NIH Grant R01NS036976

  NIH · $744k · 2005

• The representation of uncertainty in the sensorimotor system

  NIH · $1.6M · 2011–2018

• Development of a Bidirectional Brain Machine Interface

  NIH · $6.7M · 2005–2016

• NIH Grant R01NS043323

  NIH · $957k · 2008

• A primate model of an intra-cortically controlled FES prosthesis for grasp

  NIH · $7.4M · 2006–2023

Frequent coauthors

• Matthew G. Perich

  58 shared

• Chethan Pandarinath

  The Wallace H. Coulter Department of Biomedical Engineering

  51 shared

• Raeed H. Chowdhury

  Center for the Neural Basis of Cognition

  47 shared

• Konrad P. Körding

  44 shared

• Juan A. Gallego

  Imperial College London

  40 shared

• Sliman J. Bensmaı̈a

  36 shared

• Christopher Versteeg

  32 shared

• Eric J. Perreault

  Shirley Ryan AbilityLab

  31 shared

Labs

• Miller LabPI