Product Update - April 2018
We’re now deep into April, and progress is zipping along. Since we’re between iterations this time, this will be a more digestible (read: shorter) post.
Jacob Flood, March 19, 2018
This week's got progress on the latest engineering validation prototype, as well as the software test's we're running. Enjoy!
Quick note: we recently switched email service providers - as a result, we've had a few glitches with email bounces. If you haven't gotten a reply from an email you've sent to us, this is likely why. We're sorry for the issues - send your message again, and we'll get to it!
The EV2 iteration is coming along great – we’ll have a full breakdown in our next update, along with details about the tooling. On software, we’re working on our electrode evaluation test, and our latest UX-centric data acquisition. Everything is still on track for September.
We have some cool preliminary results from our clinical study, as well as some info on our larger mission at the end of the post!
We mentioned last month that we’re working on a second iteration of our engineering validation prototype (EV2). We’re on schedule with this iteration, and things are going well.
This version has several changes from the previous one, namely:
- New electrode mechanism, which we described in detail last month
- New electrodes, with better material than last version
- A better shape, motivated by the anthropomorphic analysis we performed
- Better ear cup isolation, following from our first audio tuning edits
- Modified cable harness (what holds all the wires together)
- Various CMF changes to the logo disk, ear cup fabric, and correcting for color mismatch
- Addition of ANC, which was not active in EV1
- Adjustment of the connector for fabric electrodes
- Adjustment of internal cavity volume of cup, for sound quality
- Adjustment of ear cavity shape for comfort, in order to fit all ears
- Improved switching between Bluetooth & wired audio
- Made electrodes removable
- Merged top band PCBs to cut number of cables passing from one cup to another. By extension, this makes the mounting more robust
- Grounding & shielding optimization, for signal quality
EV2 incorporates lots of subtle changes – by making these decisions before moving to tooling, we’re saving the time and cost of having to retool any of the parts. EV2 will become the gold standard against which we measure the quality of each of the mass production units.
David and Chris have been in China this whole month working on the iteration – we can’t wait to finish it and start testing it out with our software back in Montreal. We’ll be doing an analysis of the build this month, so we’ll wait until next update to give a full breakdown of what works and doesn’t in this iteration. Our plan is still to move to tooling after this iteration, and from our preliminary checks we’re in-line with this plan.
In parallel, the software team has been working hard on two projects: the EET, and the NASDAQ. Naturally, we love acronyms.
The EET is our Electrode Evaluation Test: a protocol we can run to determine the quality of the entire electrode stack. This, it turns out, is a difficult problem: there’s no single standardized test for measuring electrode quality.
Up until now, our process for making high quality electrodes has been to separate the problem, and optimize each part separately. The material, the skin contact, the analog electronics, and the digital conversion were each optimized separately, in order to ensure that they hit the noise requirements we set for the system – each part of the stack had a different requirement.
As we assembled the stack, optimizations related to ground loops, wire quality, solder type, and assembly process became more important – these were designed using the noise metrics from the previous step. Now that the entire stack is complete, the last step is to measure the quality on a final metric – something that indicates the overall ability for our electrodes to measure brain activity, as compared to medical grade sensors.
What we found was that there’s a lot of debate in the industry as to how this final test should take place. Should we have users close their eyes and measure an alpha rhythm as a metric of brain activity, or flash a light and measure a steady state evoked potential? Or should we nix all of those, and simply compare the second-by-second voltage change between our electrodes and a medical grade standard? There’s no single test that perfectly encapsulates the signal quality.
Our neuroengineering lead has been hard at work determining the best way to evaluate this process – we’ll have more information on the results we get in our next update.
In parallel, we’ve been working on the NASDAQ: the Neuro-Adaptive System Data Acquisition. In this experiment we’ll be sitting users down using our latest Mindset, running a stripped version of our full app, in order to test the experience of labelling data within the app.
A big priority with the app experience is to ensure that the infrastructure is in place to allow our algorithms to get better over time. In order for that to occur, we need to embed the ability for you, the users, to tell the system when it gets a prediction right or wrong.
This is fundamental to machine learning systems: in order to get better, they need to be given feedback. YouTube, for example, bakes this into the experience: whenever it suggests videos to you, it tracks which ones you click, and uses that to learn what went right and what went wrong in its prediction. This way, over time, the suggestion engine gets better at recognizing the patterns that indicate what video you want to see next.
Our system works the same way: over time, you’ll be able to train Mindset to recognize different neural states. We’re starting with concentration, but soon we’ll be including fatigue, motivation, stress, and other high-level neural states. This is a really exciting part of the experience: the more you use Mindset, the more accurate it will get, and the more value you’ll see in tracking your states over time.
In order for this to work, we need to bake the training experience into the product, like YouTube. The NASDAQ will test out the protocols we’ve designed for getting user feedback on the prediction of our system, in a subtle, behind-the-scenes way. We’ll have more information on this process (and on the app in general) after we run this experiment!
SCIENCE, AND MORE
Separate from the product development, a team out of the Montreal Neurological Institute has been working with us to run experiments related to attention, mind wandering, and different physiological measures. These studies contribute valuable data to us, which we use to build our prediction models, but also help to advance the scientific community’s understanding of the brain.
One such experiment we’re running is looking at the correlates of physiological measures (heart rate, galvanic skin response), neural measures (EEG sensing), and personality traits (Big 5 personality test) to reliable measures of attention, creativity, mind wandering, and other psychological outputs. The results are still coming in, so we can’t draw premature conclusions, but we’re excited about the finding. One fun tidbit: we have preliminary results that indicate a correlation between openness to experiences and mind wandering – shout out to anyone that can explain this phenomenon!
Product aside, this is a big part of why we’re building Mindset. Right now, our understanding of the brain is limited by our ability to measure what’s going on – in academia, the only way to get any visibility into the brain is through an fMRI (which costs millions of dollars, and takes hours to set up) or an EEG ($20,000, requiring conductive paste, and roughly 45 minutes to set up). Both are expensive, long to set up, and incredibly uncomfortable to wear – overall, not conducive to measuring real neural states day-to-day.
With Mindset, anyone can pick up and wear a low-cost brain-sensing device for hours in a row. This opens up the opportunity to measure EEG in environments never before possible – classrooms, offices, esports competitions, and more. Adrien’s research before joining Mindset focused on intentional versus non-intentional mind wandering: being able to use hundreds of EEG devices in the field to measure this phenomenon would have made his work infinitely easier. Mindset has the opportunity to facilitate significant progress in our understanding of the brain.
While this is not directly relevant to our development, we wanted to share the progress we’re making in this domain. We believe that advancements in neuroscience are going to define the future of society in a significant way – we’re honored to be working with experts on the cutting edge of this research, to help move the bar in a meaningful way.
We’ll finish with a little teaser. We’ve been working behind the scenes on building partnerships that will help us grow both the product and the company. We recently found several partners that are very interested in working with us on the next steps for Mindset. While we can’t say anything publicly yet, we’re really excited about the added value these partners will bring over the next months. Stay tuned for more!
As we get closer to shipping, we’ll be needing more and more help from you, our backers, in order to motivate product and marketing decisions. Do you prefer additional software features, or more hardware functions? Do you want applications in education, or in gaming? All of these are questions we’re debating, that we would love feedback on.
To help guide this, we created a private Facebook group, where we’ll ask questions, and you can contribute ideas towards how to improve our products. This will be separate from our normal page, exclusively for backers to contribute to the development of the product. You can join the group here. Look forward to hearing from you!
That’s all folks! Leave us a comment below with any questions or thoughts you have on this update, and we’ll be sure to get to it.
As always, lots of love!
- The Mindset Team