The third issue of the astrocyte* newsletter is here. Among other things, it covers:

Neurotech reading list
- MEG - scanning while moving, how quantum tech improves MEG, and MEG vs. fMRI;
- tools - preprocessing/processing pipelines for EEG, fMRI;
- improving EEG with data science - a plugin for NNs to handle missing EEG channels, a CNN for solving the EEG inverse problem in a distributed dipole model.
Startup/Corporate news
a) Investments: two TMS companies raised capital, big rounds for Synchron and Precision Neuroscience;
b) Other news: the CE Mark for a BCI software, FDA 510(k) Clearance for a minimally invasive cEEG system.

My interpretations/comments are in italic.

📬📬📬 Subscribe to get a monthly update on neurotech and neurotech investment/commercialisation activity in your mailbox.

Subscribe now

Thanks!

Peter

***

I. Reading List

Tech Stack

Using a 90 channel (45-sensor) whole-head OP-MEG system and concurrent motion capture, we recorded auditory evoked fields while participants were… standing up and making large natural head movements continuously throughout the recording.

There are already MRI scanners with lower maximum field strengths that do not require a protected room or building. Although these scanners have limited resolution quality of resultant MRI images due to a lower signal-to-noise ratio, they have a market.
OP-MEG systems may move in a similar direction when systems that don’t require shielding will be created for some use cases where a lower resolution works just fine.

$Twitter avatar for @neurofractal$

Robert Seymour @neurofractal

Pleased to share our new pre-print from the UCL OPM team and the Maguire Lab! @Nic_AAlexander @ofluff @WCHN_UCL Using a very simple experimental design, we show that OPMs can be used to measure neural activity in standing, mobile participants

biorxiv.orgUsing OPMs to measure neural activity in standing, mobile participantsOptically pumped magnetometer-based magnetoencephalography (OP-MEG) can be used to measure neuromagnetic fields while participants move in a magnetically shielded room. Head movements in previous OP-MEG studies have been up to 20 cm translation and ∼30° rotation in a sitting position. While this rep…

Bonus - a thread on the MEG and fMRI:

Micah Allen @micahgallen

Here is my bold prediction for neuroimaging: fMRI is the past, OPM is the future. The relatively low cost, customizability, high SNR, and excellent spatiotemporal resolution will see OPMs and other MEG sensors emerge as the dominant functional imaging tech in the next decade.

Quantum physics gives brain-sensing MEG scanners a boost - … the development of optically pumped magnetometers (OPMs) has been key. These are quantum enabled magnetic field sensors that offer similar sensitivity to SQUIDs but without the need for cryogenics. … OPMs can be mounted directly on the surface of a human head, increasing sensitivity by removing the thermally insulating gap and getting the sensor closer to the brain. This also allows the sensor array to move with the head, making the MEG measurement resilient to subject motion.

I’d advise neurotech entrepreneurs to carefully monitor adjacent fields and prepare to apply to neurotech what were achieved there.

On OP-MEG and speech tracking:

Matthew Brookes @MattBrookesMEG

Absolutely delighted to see this published in @NeuroImage_EiC !! Cortical tracking of speech measured using OPM-MEG! Wonderful collaboration with Mathieu Bourguignon!! Turns out you can do low frequencies using OPMs!! @UoN_MEG @CercaMagnetics sciencedirect.com/science/articl…

Brain Modelling as a Service: The Virtual Brain on EBRAINS. The Virtual Brain (TVB) is now available as open-source cloud ecosystem on EBRAINS, a shared digital research platform for brain science.

There’s always a debate about what to build first - applications or tools. It’s fantastic that projects like this offer tools to researchers/entrepreneurs (here) and allow them to focus on doing research/building applications.

Petra Ritter @_PetraRitter

Brain Modelling as a Service: The Virtual Brain on EBRAINS arxiv.org/abs/2102.05888 - Fantastic team work! @HumanBrainProj @EBRAINS_eu @pswieboda @EOSC_eu @berlinnovation @ChariteBerlin @Fenix_RI_eu @juniquefr @ECDigitalFuture @BIDSstandard @ECN_Berlin @bccn_berlin @TVB_cloud

Data Science

Google and Harvard map brain connections in unprecedented detail. The end result, which Google calls the H01 dataset, is one of the most comprehensive maps of the human brain ever compiled. It contains 50,000 cells and 130 million synapses, as well as smaller segments of the cells such axons, dendrites, myelin and cilia.

In the recent neurotech landscape, I highlighted that neurotech startups feel less pressure from tech giants than ‘traditional’ software companies that fearlessly compete with them. However, tech giants are building neuro datasets and at some point, may increase their activity in the market. Therefore, those who consider launching a neurotech startup should probably move faster.

In this article, we present SynthSR, a method to train a CNN that receives one or more [MRI] scans with spaced slices, acquired with different contrast, resolution and orientation, and produces an isotropic scan of canonical contrast (typically a 1 mm MP-RAGE).

Juan Eugenio Iglesias @JuanEugenioIgl1

Our paper on joint super-resolution & synthesis is finally published in NeuroImage. Turn one or multiple clinical scans from a subject (of any resolution/ orientation/ contrast) into a 1mm MPRAGE you can easily segment, register... (eg with @FreeSurferMRI)

sciencedirect.comJoint super-resolution and synthesis of 1 mm isotropic MP-RAGE volumes from clinical MRI exams with scans of different orientation, resolut…Most existing algorithms for automatic 3D morphometry of human brain MRI scans are designed for data with near-isotropic voxels at approximately 1 mm …

In fMRI, researchers have many degrees of freedom in the way that they can process the data… Here we show, using three popular whole-brain dynamical models, that different choices during fMRI preprocessing can dramatically affect model fits and interpretations of findings.

Stewart Heitmann @StewartHeitmann

Another paper using the #BrainDynamicsToolbox. This one by @Kevin_M_Aquino on the intersection between data quality and dynamical modelling of large-scale fMRI signals

biorxiv.orgOn the intersection between data quality and dynamical modelling of large-scale fMRI signalsLarge-scale dynamics of the brain are routinely modelled using systems of nonlinear dynamical equations that describe the evolution of population-level activity, with distinct neural populations often coupled according to an empirically measured structural connection matrix. This modelling approach …

EPOS: EEG Processing Open-Source Scripts. Since the replication crisis, standardization has become even more important in psychological science and neuroscience. As a result, many methods are being reconsidered, and researchers’ degrees of freedom in these methods are being discussed as a potential source of inconsistencies across studies. With the aim of addressing these subjectivity issues, we have been working on a tutorial-like EEG (pre-)processing pipeline to achieve an automated method…

Something built for research purposes may transform/be plugged into a developer tool that fuels neurotech adoption.
Positive for reproducibility, therefore for wider neurotech adoption.

Martin Weiß @mrtnwss

Happy to share our new methods paper including open-source (pre-)processing scripts for EEG research published in @FrontNeurosci! doi.org/10.3389/fnins.… 🧠@AcademicEEG

doi.orgEPOS: EEG Processing Open-Source ScriptsBackgroundSince the replication crisis, standardization has become even more important in psychological science and neuroscience. As a result, many methods are being reconsidered, and researchers’ degrees of freedom in these methods are being discussed as a potential source of inconsistencies across…

Robust learning from corrupted EEG with dynamic spatial filtering. We propose dynamic spatial filtering (DSF), a multi-head attention module that can be plugged in before the first layer of a neural network to handle missing EEG channels by learning to focus on good channels and to ignore bad ones.

It’s an attempt to overcome the limited computing power of consumer-grade EEG devices that may boost adoption.

Hubert Banville @hubertjbanville

How should we train predictive models on corrupted #EEG? Check out our new preprint on Dynamic Spatial Filtering arxiv.org/abs/2105.12916 - joint work with @seanunwood @aimone @dngman @agramfort. We use attention & data augmentation to train robust models on sparse EEG. Thread:

This figure shows an example of the channel corruption process with additive white noise on a window of six channel EEG, along with a visualization of the distribution of channel importance for each channel as channels T3 and T4 are sequentially corrupted.

ConvDip: A Convolutional Neural Network for Better EEG Source Imaging. Artificial neural networks have been previously used successfully to find either one or two dipole sources. We present ConvDip, a novel convolutional neural network (CNN) architecture, that solves the EEG inverse problem in a distributed dipole model based on simulated EEG data.

Tomas Ros @neuromeditate

Deep learning for EEG source-localisation! ConvDip: A convolutional neural network for better M/EEG Source Imaging biorxiv.org/content/10.110…

Other Themes

A brain-computer interface that evokes tactile sensations improves robotic arm control. We supplemented vision with tactile percepts evoked using a bidirectional brain-computer interface that records neural activity from the motor cortex and generates tactile sensations through intracortical microstimulation of the somatosensory cortex.

I’m a big believer in bidirectional BCIs that allow personalisation for stimulation and other purposes. Personalisation is a powerful driver of consumer/enterprise tech and should be emphasised in neurotech too.

Sharlene Flesher @WishUponAShar

“Touch is hard to grasp” is the best play on words for this paper that I have ever seen

RNEL @RNELabs

Touch is special! It makes us feel things, and move better. But touch is hard to grasp! That’s why we worked a lot to deliver sensory feedback to the cortex of person with SCI and integrate it into a Brain controlled robot. #TouchMakesUsSpecial 💛💙🦾 https://t.co/g01piHCQL1 https://t.co/sxCENfTf6b

II. Startup/Corporate News

Investments

Precision Neuroscience Raises $12M to Develop Next Generation Brain-Computer Interface Technology (🇺🇸). A very minimalistic website tells us - ‘Our integrated platform – delivery, device, and data – will help deliver the next frontier of medicine’.
I’d highlight the ‘integrated platform’ part, and it seems that multiple startups aim to build ‘the Apple of neurotech’.
Synchron Secures $40M in Series B led by Khosla Ventures to Launch U.S. Clinical Trials of Minimally Invasive Brain Computer Interface (🇺🇸/🇦🇺).
Synchron is building a minimally invasive system, and therefore may have a wider adoption, probably by millions of people. Compare it to ‘traditional’ DBS systems that were implanted to 200K+ patients, according to some estimates.
Alex Morgan, MD PhD @genomicsdoc
Our newest portfolio company is using smart stents to measure electrical activity in the brain through the cerebrovascular system, the natural highways of the brain.
Thomas Oxley @tomoxl
Khosla Ventures agree - the future is minimally invasive brain computer interfaces. Very excited to ramp up towards a US clinical trial for our first indication in paralysis. Disruptive endovascular neuromodulation pipeline to follow. #BCI https://t.co/fheGSZPuSd
7:00 PM ∙ Jun 3, 2021
27Likes5Retweets
Actipulse Neuroscience (🇺🇸) that develops a neuromodulation treatment for depression, launched a crowdfunding campaign. I’d encourage you to dive into their prospectus for some exciting details, for example:
1. The company is profitable, raised only $140K from investors, but has already 210+ hospital-setting neuromodulation devices currently in use;
2. Business model: 1) patient direct use with our at home device (post FDA approval); 2) shared-revenue for our hospital-setting device.
Greenbrook TMS Announces Completion of US$23.5 Million Private Placement (🇺🇸). The company runs outpatient mental health service centres and provides magnetic stimulation therapy.
It’s not a neurotech company per se, but learning about it could help neurotech entrepreneurs to gauge the scale of an opportunity. Greenbrook provided more than 510,000 TMS treatments to over 14,000 patients. According to Pitchbook, a database, Greenbrook’s revenues grew from $6.7M in 2016 to $43M in 2020.

Trials & Other Announcements

The Wyss Center announces CE mark for its brain-computer interface software for communication, NeuroKey BCI (🇨🇭). NeuroKey BCI is currently in home-use as part of a single case study to enable communication for a user who is completely locked-in because of ALS.
iCE Neurosystems Announces FDA 510(k) Clearance for Brain Monitoring Devices Used in Critical Illness (🇺🇸). Devices are part of the iCEWav Neuromonitoring Platform which combines minimally invasive electrodes, integrated multiparametric data and custom cloud software to drive AI-based clinical decision support.

📬📬📬 Subscribe to get a monthly update on neurotech and neurotech investment/commercialisation activity in your mailbox.

Subscribe now

astrocyte*

Neurotech Monthly / June 2021