Thread by @robjcoops

I am contemplating live-tweeting the @KernelCo launch of their TD #fNIRS device 'FLOW' - which starts in 6 minutes here:

Bit of background - I have worked in #biomedicaloptics and specifically #fNIRS for 10 years. I remember one of my PhD supervisors telling me there was no research career to be made in diffuse optics. I disagreed

My particular goal over the last few years has been to try to be the first to demonstrate high-density, whole-scalp, wearable diffuse optical imaging (DOT)/#fNIRS. @gowerlabs and @UCL_DOTHUB made a lot of progress towards that

Now @KernelCo comes along! But hey, exciting for the field none-the-less.

Actually myself and Jem Hebden and friends at the University of Edinburgh wrote a grant in 2016 to @EPSRC to create a wearable TD-NIRS module using integrated circuitry and SPADS. Basically what @KernelCo seem to have done. It was rejected. Reviewers said that it was pointless...

Because HD-DOT (CW) was already good. 🤐🤯

Seems to be starting!

I am opening a beer. Key question: how are they achieving their claimed 200Hz sampling rate and still getting sufficient SNR? #fNIRS

By the way the device 'FLOW' does not measure blood flow.

Surprisingly amateurish presentation for someone who just raised $53 Million dollars for this

It's quite large. #kernal #fNIRS

It's Katherine Perdue! Yeah- GO KATHERINE

Great presentation! Not sure why is she in a greenhouse?

Looks like one detector and six sources per module? Or vice-versa? #fNIRS

It's Joy Hirsch! And now @IliasTachtsidis from @UCL Wow - lot's of familiar faces. #fNIRS

I still have no idea where the business model is here. Could anyone explain to me? #kernel #neuroscience

@BoasDavid popping up too! Is literally everyone in @SfNIRS in this video? 😂

Here we go - a bit of tech now #fNIRS #kernel with Ryan Field. A quick reminder from me, that @kernel started with a plan to build implantable devices, like @neuralink then pivoted away from an invasive approach.

Looks like a nice SPAD array there. Four channels apparently. The Gamma version is a single channel- that's what's in the system. #fNIRS

On to source side. Pulsed laser system - unclear what exact technology is. Thank you Ryan! Impressive! #fNIRS

That's PROFESSOR David Boas to you, Ryan! @BoasDavid #fNIRS

52 hexagonal modules. Single source, six detector. That makes sense actually, for timing reasons- sources cost you time - detectors do not. #fNIRS #kernal

Unclear if the modules form channels across modules - though that is not as important for time-domain as you can gate it and do null-separations. #fNIRS

Sounds like YES - between modules. So 10mm channels on-module, and then longer across module. Get out to ~35mm source-detector distance. #fNIRS

Interesting hub-spoke microcontroller/module configuration. #fNIRS. Nice TPSFs in real time!

200Hz TPSFs is damn awesome! #fNIRS @KernelCo - accepting 800 million cps - that's genuinely impressive. MONSTIR could do about 1 million. Pile up? Good question @BoasDavid

So that makes the dynamic range issue a lot easier, because you are not going to saturate at close channels. Five decades dynamic range is impressive. Background light issues somewhat overcome by the cps too. #kernal #fNIRS

Oooh. Integrated #EEG - that's cool #kernel #fNIRS

Yey SNIRF! Well done @SfNIRS @BoasDavid

So @KernelCo are giving away 10x 50 module devices on the basis of the scientific and market potential of the application, and then 40x for $5000

No particular depth sensitivity, no particular improvements in resolution. 30Hz whole head. That's too slow for #EROS but unnecessarily fast for #fNIRS. Good for SNR I guess. Also good they are working with the protocols developed by the TD-NIRS academic community.

I'm not convinced the spatial resolution of the @KernelCo FLOW device will be better than @gowerlabs LUMO, despite the fact they are using TD. Fare lower density of sources.

What is this brain data showing? No scale, no task? I don't get what this is? Nice music though. #fNIRS #kernel

Whole scalp ~= whole brain. #fNIRS #kernel

That was my question! Is it wired. WOOOP #fNIRS #kernel

Okay - it's over. Take home points?
1) I was amazed how much time was spent on my friends in the #fNIRS community, but not many neuroscientists. Not much clarity on who the user is and what the vision is.

2) The CPS and therefore the sample rate are genuinely impressive. 100-1000x more counts per second are measurable. Why does that matter? It means you won't saturate, so you can have shorter channels without filtering. what was less clear was sensitivity at the low end

it also might buy you out of background light issues, but unclear what happens if background light levels are changing- like almost all halogen and LED lights. Mean time might be robust, but total counts won't be.

3) If you can't sample fast enough for/don't believe in #EROS, then why not just downsample everything and get better SNR? #fNIRS

4) the mechanics is suboptimal - unclear why the modules have to be so tall - obviously spring-loaded with hair in mind, but that makes studies of children/infants much harder. Can we remove the sprung mechanics? #fNIRS

Anyhoo. That was very cool. Congrats to @KernelCo and my friend Katherine. Can I have one please? 😍

@threadreaderapp unroll