October 27, 2020 | Bryan Johnson
Kernel Flow 50 Live Stream Summary (tl;dw)
Last week we announced that the first fifty Kernel Flow systems will become available for select partners starting in January 2021 (application). We live streamed a sixty minute event explaining the technology and hearing from experts. We also posted online the full tear down video.
Why Kernel Flow Could Represent a Step Function Change
Upon learning certain details of what we’ve built, NIRS pioneer and expert Dr. David Boas remarked “that’s insane.” That has been repeated by many people in the field. Here are some of the advancements they are considering when making that statement:
- Kernel Flow captures 1,000x more information than existing TD-fNIRS systems. Flow collects 1 billion photons/sec, per detector. Current state-of-the-art TD-fNIRS systems traffic jam at around 1 million photons/ sec. per detector.
- 200 Hz sampling frequency which is 200x faster than most systems on the market today. The faster sampling frequency will enable, among other things, a) higher resolution hemodynamic data b) more precise motion artifact removal and c) better reduction of physiological noise.
- Whole head coverage. Most TD-fNIRS systems in use today cover only a fraction of the head -- mostly due to the high prices. Flow covers the whole head (with 312 detectors, 1,000+ channels). Whole-head coverage allows study of cortical activity as a network, an approach which has gained much traction in recent years in the fMRI literature (the functional “connectome”, “network neuroscience”).
- Absolute measurements. fMRI can only measure relative changes in blood oxygenation. Flow, powered by time-domain spectroscopy, will offer reconstruction of absolute concentrations of oxygenated and deoxygenated hemoglobin, in 3D throughout the cortex (currently a work in progress). No non-invasive brain interface in the world currently offers this kind of absolute measurements with full head coverage. When I measure my blood glucose levels or cholesterol (especially across days), I want an absolute measurement, not relative. It’s TBD how valuable absolute reconstruction will be but to me it's one of the most promising enhancements.
- Flow will generate the richest neuroscience datasets in history. Due to the scaling power of Flow (cost, manufacturing, ease of use), we will build the world’s richest neuroscience datasets, in record time: longitudinal measurements from individuals and groups, in labeled naturalistic environments, with social interaction. Novel experimental protocols and analysis strategies will likely dovetail with this expanded data context to foster accelerated insights.
- Flow will ship with a processing and analysis API and sandbox, powered by cloud infrastructure, that will allow users of any technical background to quickly plug and play with state-of-the-art methodologies.
We have dramatically changed the game in multiple areas simultaneously - significant cost reduction, ease of use, neural signal quality (1,000x more information), sampling frequency (200x the speed), whole head coverage, absolute measurements, labeled naturalistic environments, longitudinal abilities, human-to-human interactions, the largest datasets in history, a single standard of calibration across all Flow devices and a greatly expanded ecosystem of people who are now empowered to interact with the brain and mind.
It is difficult to model and predict what becomes possible when all of these things are enabled at the exact same time.
So what else are you capable of? The Kernel Flow 50 program is an effort to begin answering this question. We will soon find out.
What Others Are Saying
Both of these technological advances—time domain and full-head coverage—are essential. They're essential because the brain doesn't work in isolated units. It works as an integrated whole.
- Joy Hirsch, Professor at Yale University, Psychiatry, Comparative Medicine, Neuroscience
Giving access to this equipment to more people—to more scientists, psychologists, clinicians—it opens up a lot of diversity to what can be investigated.
- Ilias Tachtsidis, Professor at University College London, Biomedical Engineering
Well, that's impressive. I mean with all that, just to fit in this small footprint. I think it's amazing.
- Hasan Ayaz, Professor at Drexel University, Biomedical Engineering
- Dr. David Boas, Director of Neurophotonics, Boston University
Fun livestream by @kernelco today; unvarnished and informative. Some hard engineering to get an impressive fNIRS helmet with 200 Hz sampling, and now they want the community to help them figure out what it’s good for. “VGA fMRI, on the fly.”
- Philip Sabes, Emeritus Professor of Physiology, UCSF
Kernel Flow Specs
- A modular TD-fNIRS system available in configurations from 1 to 8 plates of modules.
- 8 plates of modules provide full coverage for an average size head.
- Each plate contains 4-8 modules
- Each module has one dual-wavelength source (690nm & 850nm) and 6 TD detectors.
- Whole head coverage includes 52 modules (52 sources, 312 detectors, 1,000+ channels, since each detector can receive light from several sources)
- All modules are synchronized across the system.
- 200 Hz sampling frequency
- Integrated EEG with up to 8 active dry electrodes
- Dedicated task acquisition PC
- Auxiliary device input for syncing external hardware, connects over USB to acquisition PC
- Each of the 1000+ channels captures the histogram of photon arrival times after having traveled through the brain, providing much richer and deeper information about its optical properties than traditional fNIRS systems
- The time domain information also enables extraction of absolute blood oxygenation throughout the cortex (as opposed to relative measures like fMRI), providing a more precise quantification of neural activity than any other hemodynamic measurement.
- Access to the Kernel API, development sandbox, and cloud-based data processing and analysis infrastructure.
- Export data to standard .snirf format for offline processing.
- Full online documentation
- Robust process checking to avoid mistakes - built in automated quality controls
- Secure data storage
- Study tracking dashboard for easy and safe monitoring of participant progress
- Reference task codebase for easy integration with your existing workflows
- Daily diagnostics/logging with preventative support - identify problems before they become critical. Our goal is to help you achieve 100% uptime and 0 avoidable mistakes.
- Community Support Knowledge Base
How To Apply
Submissions to the Flow 50 program will be accepted from Nov. 1, 2020 through December 31, 2020. System allocations will be granted on a rolling basis. Here is the application.