Contact

Overview

The Neural Engineering Group sits at the intersection of neuroscience, electrical engineering, and biomedical materials. Its mission is to build the devices, algorithms, and analytical frameworks needed to read from and write to neural circuits with the precision and temporal resolution required for both basic research and therapeutic applications. The group's portfolio spans electrode array design, ASIC-level signal conditioning, real-time spike-sorting, and closed-loop control of neural stimulation based on decoded neural state.

On the device side, the group fabricates flexible polyimide electrode arrays with up to 512 channels and develops hydrogel coatings that reduce the chronic foreign-body response, extending implant functional lifetime. On the algorithm side, it applies deep learning-based decoders for motor intention, and model-predictive control frameworks for adaptive deep-brain stimulation protocols. A dedicated cleanroom bay in the Cleanroom & Nanofabrication Facility is reserved for group device fabrication.

The group collaborates with the Computational Neuroscience Group on biophysical models that guide stimulation parameter selection, and with the Perception & Decision Lab on BCI protocols that exploit decision-related neural dynamics. Its clinical translation work is pursued through a collaborative agreement with the Veyra Institute's commercial consulting arm.

Research themes

  • High-density, flexible electrode arrays for chronic cortical and subcortical recording
  • ASIC design for low-noise, low-power multichannel neural front-ends
  • Real-time spike sorting and field potential decomposition on FPGA and edge hardware
  • Deep learning decoders for motor imagery and speech-based brain-computer interfaces
  • Closed-loop adaptive deep-brain stimulation for movement disorders
  • Biocompatible materials and surface chemistry for long-term implant stability

Current projects

Active research programmes, 2024–2027

Project · VX-NEG-01

FlexNet: 512-Channel Flexible Cortical Array

Fabricating next-generation polyimide electrode arrays with 512 channels, 12 µm inter-electrode pitch, and integrated multiplexing to reduce tethering wire count. Hydrogel encapsulation is optimised for a 24-month functional lifetime in vivo in a non-human primate model.

Funding: VIAS Research Excellence Grant · 620,000 cr

Project · VX-NEG-02

DeepDecode: Neural Decoder for Motor BCI

Training convolutional-recurrent neural networks on motor cortex population activity to decode intended limb trajectory from 150-channel recordings in real time at <20 ms latency. Benchmarks against Kalman filter and linear discriminant analysis baselines using a standardised BCI challenge dataset.

Funding: External industry contract VX-BCI-24 · 440,000 cr

Project · VX-NEG-03

AdaptDBS: Closed-Loop Deep Brain Stimulation

Implementing model-predictive control of subthalamic nucleus stimulation in a Parkinson's disease patient model, with stimulation parameters updated every 250 ms from decoded local field potential biomarkers. Aims to reduce stimulation energy by 40% relative to open-loop protocols while maintaining therapeutic efficacy.

Funding: VIAS Translational Research Fund · 510,000 cr

Selected publications

  • Halmstad L., Okeke B., Schwarz P. "512-channel polyimide cortical array with integrated multiplexing: fabrication and acute recording quality." VIAS Journal of Neural Engineering 21(3), 2024. DOI: 10.veyra/VX-4462
  • Halmstad L., Tran V. "Convolutional-recurrent decoders outperform linear methods for hand trajectory prediction from M1 ensembles." Veyra Brain-Computer Interfaces 7, 2023. DOI: 10.veyra/VX-4228
  • Schwarz P., Halmstad L. "Zwitterionic hydrogel coatings suppress astrocyte encapsulation of cortical implants for 18 months." VIAS Biomaterials 14(2), 2023. DOI: 10.veyra/VX-4079
  • Okeke B., Halmstad L., Steiner R. "Decoding decision confidence from fronto-parietal LFP for closed-loop BCI gating." Veyra Neural Dynamics 5(3), 2022. DOI: 10.veyra/VX-3903
  • Tran V., Halmstad L. "Real-time spike sorting on FPGA using online k-means with automatic cluster number selection." VIAS Engineering in Medicine 9, 2021. DOI: 10.veyra/VX-3672
  • Halmstad L. "Model-predictive closed-loop DBS: simulating subthalamic nucleus control under beta-band feedback." VIAS Computational Medicine 2(1), 2020. DOI: 10.veyra/VX-3449

People

Group lead: Dr. Lior Halmstad · View all Veyra people

Postdoctoral researchers: Dr. Benjamin Okeke, Dr. Philipp Schwarz, Dr. Van Tran, Dr. Astrid Moe.
PhD students: Yosef Matta, Siobhan Daly, Cédric Lamarre, Priya Raghunathan, Karan Singhania, Dina Osei, Tomáš Blaho, Ingrid Holmberg.
Research staff: Ivan Sorokín (FPGA engineer), Ngozi Anyanwu (device fabrication), Mihail Gavrilov (cleanroom specialist).

Related links