Neurotech News #1: Hello World

Research Round Up: Recent Papers That Caught Our Interest

^{by Robert Murcko, Yuqing Wang, Holland Ernst}

The human subthalamic nucleus transiently inhibits active attentional processes [Accepted Manuscript] (Soh et al., March 2024)

• The subthalamic nucleus (STN) is known to be important for the inhibitory control of motor activity. There is a hypothesis that the STN can also inhibit non-motor related activity and that deep brain stimulation of that region would affect that non-motor inhibition response. This hypothesis was tested in human Parkinson’s Disease patients with existing DBS implants in the STN region.

• Two separate experiments were conducted within this study and there was 0% overlap in the participant samples.

• Experiment 1

  • Intracranial local field potentials (LFPs) were recorded from DBS implants in the participant’s STN brain regions during a set of visual attention tasks. High density EEGs (HD EEGs) were also used for pinpointing steady-state visually evoked potentials (SSVEPs) during these tasks. This experiment played surprising and distracting sounds during the visual task and then tracked and analyzed the biosignals above to quantitatively determine if STN activity correlates to inhibition of SSVEP.

  • Results of experiment 1 showed gamma frequencies (greater than 60 Hz) in the STN preceding a drop in SSVEP after the distracting sounds were played. Analysis performed across trials suggested with statistical significance that the gamma activity in the STN moderated the effect of distracting sounds on the SSVEPs.

    

    Figure 3A (Soh et al., 2024)

• Experiment 2

  • In the second experiment, STN activity was modulated via the DBS implant across two separate study sessions of the same task from experiment 1. The intent of this experiment was to run a causal test of the relationship between STN activity and non-motor inhibition.

  • Results of experiment 2 showed that when the STN is stimulated via DBS the SSVEP reductions caused by distracting sounds were minimized. This is evidence that STN plays a role in attentional-inhibition in moments of surprise.

• Importance: This work provides evidence for the theory that STN DBS treatment affects cognitive processes and functioning. It is also an important example of rigorous research conducted using the subcortical LFPs from implanted DBS devices in an outpatient setting.  

• Next Steps: Next steps may include investigation of STN function in working memory or other non-motor functions. It may also include further exploration of the possible cognitive side effects of STN DBS.

Customizing the human-avatar mapping based on EEG error related potentials (Iwane et al., March 2024)

• Via ErrPs in EEG recording, the novel VR-integrated BCI detects breaks-in-embodiment (BiE) to enable auto-adjustment of the human-avatar mapping. 

• Results demonstrate successful prediction and customization of mapping using BCI-reinforcement learning without interruption to the user’s experience. The system maintains user engagement and immersive experiences in VR while detecting and correcting issues. 

• Participants’ perception of distortion and occurrence of BiE increased progressively with the magnitude of distortion introduced in the virtual environment. However, both personalized and non-personalized decoders successfully detected BiE with high accuracy and the algorithm effectively determined the optimal distortion level based on users’ implicit feedback.

• Importance: This innovative approach addresses a crucial challenge in VR applications by enhancing user experience through real-time adjustment of human-avatar mapping. Implicit detection and correction of errors without interrupting user interaction allows for customization and offers significant implications for maintaining engagement and immersion in VR environments, with broader implications for commercial VR products.

• Next Steps: The authors mentioned potential future directions of the BCI-VR system over longitudinal sessions to assess the generalizability of the decoder and the user’s adaptability to visuo-proprioceptive conflicts over time. The continued advancement of BCI-VR technology holds promise for optimizing human-avatar interactions and enhancing immersive experiences in virtual environments. Further optimizing the BCI- VR system for commercial applications should be explored.

Plug-and-Play Stability for Intracortical Brain-Computer Interfaces: A One-Year Demonstration of Seamless Brain-to-Text Communication (Fan et al., February 2024)

• A method of self-calibration for communication intracortical brain-computer interfaces (iBCIs) is proposed. These systems have classically required supervised data collection for calibration.

• The proposed method uses LLMs to automatically correct output errors from the iBCI. The corrected outputs (aka pseudo-labels) are used to continuously update and calibrate the iBCI’s decoding algorithm. These updates occur online and do not inhibit the user from using the system during calibration. The framework is called Continual Online Recalibration with Pseudo-labels (CORP).  

• CORP was evaluated on one clinical trial participant over the course of 403 days. An accuracy level of 93.84% was achieved in decoding for an online handwriting task. This showed substantial improvement compared to baseline methods.

• The process that this framework follows is a bit complex, but the figure below gives a good flow chart of how the information flows. The key point is that new data (threshold crossing neural features and large language model outputs) are combined and augmented with the older versions of these same data sources. Then an  optimization algorithm called stochastic gradient descent is applied to this merged data to allow for updates to the recurrent neural network that is doing most of the “heavy lifting” in the decoding process. 

  

  Figure 1 (Fan et al., 2024)

• Importance: Because neural signals are not stationary and tend to see drift over relatively short time spans, a key problem that limits clinical translation of iBCIs is the need to constantly manually recalibrate these systems (on the time scale of every couple days). This work is one of the early steps towards removing this limitation. This work is also important because it represents the longest ever successful demonstration of iBCI stability in a human subject.  

• Next Steps: The authors mentioned several potential areas of further exploration including: combining CORP with distribution alignment methods, investigation of approaches for limiting distribution shift in machine learning algorithms, and investigating the performance of CORP in other iBCI applications (example: speech decoding). Furthermore, the investigation can be expanded to sample sizes larger than n = 1 to confirm that the improvements in iBCI stability can be effective and repeatable across participants.

Mapping dysfunctional circuits in the frontal cortex using deep brain stimulation (Hollunder et al., February 2024)

• It is critical to identify what frontal circuits/domains are responsible for specific brain (dys)functions.

  • Many brain functions originate from disrupted connections between the frontal cortex and basal ganglia. The subthalamic nucleus (STN) and striatum are two direct input nuclei to the basal ganglia. STN’s smaller size makes it a favorable stimulation target. STN simulation has been shown as effective therapy for 4 disorders (Parkinson’s disease (PD), dystonia (DYT), obsessive-compulsive disorder (OCD) and Tourette’s syndrome (TS)). Using the targeted brain areas that responded better to stimulation, we can pick out from the STN bundle, the circuits that were problematic in the first place.

• DBS treatment acts as a ‘functional lesion’ to downregulate specific neural circuits. The researchers employed the data analysis technique of DBS Sweet Spot Mapping. This method is employed as follows: Calculate a mapping of E-field magnitudes in a 3D volumetric-pixel (voxel) space, where each box in the grid stores E-field information in a 3D volume in space. Then perform a Spearman’s Rank Correlations between a) E-field magnitude and b) patient condition improvement. This results in a voxelized grid of association between entities a and b, we call the strongest association area, a ‘sweet spot’.

  

  Figure 1 (Hollunder et al., 2024)

• The analysis found that the most critical areas for each disorder are:

  • DYT: interconnections with somatosensory and primary motor cortices

  • TS: the supplementary motor areas (SMA)

  • PD: the premotor retions and SMAs

  • OCD: ventromedial prefrontal, dorsal anterior cingulate, dorsalateral prefrontal and orbitofrontal cortices

• Importance: As brain implants and stimulation devices become more prevalent, this paper is an important example of how we can make good use of the data these systems generate.

• Next Steps: This paper represents a critical step towards integrating DBS and brain connectomics as a scientific tool to further understand the relations between brain structure and disorders. 

All I See is Green: A Selection of Eye-Popping Deals

^{by Robert Murcko}

I wasn’t kidding when I said neurotech investment is trending upwards.  Companies in this vertical “raised $1.4 billion across 115 deals in 2023,” per PitchBook. And if the list below is any indication…2024 seems to be heating up across the neuro-investment landscape as well.

A selection of nearly $450M in deal flow:

• March 2024

  • SecondWave Systems receives $3M of non-dilutive funding from ARPA-H

  • ONWARD Medical Raises €20 Million

  • Forest Neurotech Receives $14 Million Funding Commitment

  • Nalu Medical Closes $85 Million Series E

  • BrainCheck Raises $15 Million

• February 2024

  • Mainstay Medical Raises $125 Million

  • Aviation Medical Raises $22 Million+ Series C

  • Samphire Neuroscience Raises $2.3 Million Pre-Seed

  • Elemind Technologies Closes $12 Million Seed Round

• January 2024

  • SPR Therapeutics Secures $85 Million in Additional Funding

  • Motif Neurotech Raises $18.75 Million Series A

  • Rune Labs Closes $12 Million Strategic Round

  • Cognito Therapeutics Raises $35M Series B Extension

In addition to the above - over the past few weeks I’ve spoken with founders and executives of several promising neurotech startups that are still looking to raise. To the investors reading this - it's never too late to get into this electrifying vertical.