Some questions about regular vCR pattern

[HackyDev]

Here are representations of vibration patterns from this paper

Regular vCR stimulation, Figure 1A, is characterized by the vCR period TCR, which sets the CR frequency fCR = 1/TCR at which individual fingertips received burst stimuli. Individual fingertips received stimuli at multiples of TCR/4 such that each fingertip received exactly one stimulus per CR period. Besides this constraint, stimuli were delivered to randomly selected fingertips. This type of CR stimulation is referred to as CR with rapidly varying sequences (CR RVS) in the literature (Zeitler and Tass, 2015). Additionally, we considered an m:n ON-OFF pattern by delivering stimuli for an ON-period of three CR periods, TCR, and paused the stimulation for an OFF period of two CR periods afterward (Lysyansky et al., 2011). A representative regular vibrotactile 3:2 ON-OFF CR RVS pattern, in the following denoted as regular vCR, is shown in Figure 1A.

Let's start discussion from regular vibrotactile 3:2 ON-OFF CR RVS pattern or regular vCR:

The questions:

1. The loop appears to be made up of 8 periods (TCR) based on the image in the paper. However, the paper mentions that the pattern is referred to as a 3:2 ON-OFF pattern, which suggests that the cycle is actually made up of 5 repeating periods. As a non-native speaker, it's possible that I may have misunderstood some of the information. What are you thoughts on this?

2. Based on the information in the paper, it seems that the stimuli are delivered to randomly selected fingertips, which suggests that the exact vibration pattern shown in the image may not be followed and that different random sequences may be generated by our code. Additionally, it's unclear whether the stimuli pattern is completely random across all iterations of the loop or if it is random within a single iteration and then repeated.

Related papers:

• Vibrotactile Coordinated Reset Stimulation for the Treatment of Neurological Diseases: Concepts and Device Specifications

[NeuroMod321]

Hi , My impression of the tech paper is that it is rather ambiguous in many areas and not clear cut - perhaps a deliberate strategy ?. Re: " 3:2 ON - OFF " pattern - that phrase by itself implies a 5 x cycle ( TCR ) pattern but the image becomes ambiguous as it depicts 8 x TCRs . To confuse that further - does this include the 1 x 100ms or 2 x 100mS ? If it is 3:2 ratio cycle what happens for subsequent cycles relative to the 2 x TCR pauses ? Since I can't think of a means to come to a definite answer - I am inclined to e.g. a) Have an option to select 5 or 8 TCRs in the data stream ? b) An option to select 1 or 2 x 100mS pauses ? c) For my trial I would go for synchronised e.g. 3/2/4/1/ then 1/2/4/3/ then 2/4/3/1/ then 2 x TCRs no action etc. The reason I am saying this is because I have been doing some work for HSP ( hereditary spastic paraplegia ) and I have discovered ORDER MATTERS . In my research e.g I have found that IF and ONLY IF the sequence 1/2/3/4/3/2/1 is followed strictly - it won't work. And 1/2/3/4/3/2/1 always works. If I change ANY of the pulse train order - it fails. i.e. the afferent neurons 'seem' to be able to "detect order ". Hope this helps ..

[RChadwick7]

Personally, I don't think the people who wrote the paper know all the answers, they simply documented something that does, or may work. As with all scientific papers, there seems to be a love of complexity, even when unnecessary. I believe the 3:2 refers to the duty cycle, where for every measure of three the vibration is on, there is a measure of 2 where it's off. I'm sure this is wrong, but as an example if each finger is vibrated for 300ms, it would be off for 200ms until the next finger. I think we can work more effectively by accepting we may make mistakes. If we were talking about replicating a drug, that would be something else. I'm fairly confident that if we get timing wrong, at worst it won't work. We may even be able to develop better patterns! I'm hoping to standardize on hardware, and perhaps a simple software framework. Then, we can all try what we think will work, based on the paper. Timing is very easy to change in software, or even from the device itself. I think it would be best to add an LCD and rotary control to make it easier to change settings. We may find different settings work better for different people.

[RChadwick7]

Hey HackyDev, would it be OK if we exchanged email addresses? Also, I saw your repo is gone.

[danteoh]

They discuss the intervals - 100ms on, 66ms rest per finger.

What i'm most curious about is that sequence repeated, or is it just showing "randomness"?

[danteoh]

I commented over on HealthUnlocked, but here's my code so far for triggering the vybronics vibrators (assuming you're going through a transistor and 3.3v rail):

https://gist.github.com/danteoh/e2053f28da48829812a2ddc548427e51

[Beartech]

I believe 3:2 is three on two off three on two off three on two off.... and for "random" is seems like they are random, but no finger can repeat in the same group. i.e. it could repeat if it was at the end of one of the "3 on" and at the beginning of the next in that 3. So fingers numbered 2-5 it could go 2435-5423-3245 but not 3345-2445 etc. And when it says that finger 2 is stimulated, it is stimulated on both hands at the same time. Not sure what "jitter" is referring to in the description. And I think that it should be as truly random as possible so even sequences like 2345 will possibly happen, but no real repeating pattern overall. I'm not sure if the random pattern inside a 3:2 sequence is something they got from DBS. The gloves were developed to do the same thing as DBS without internal electrodes. I might need to do some reading up on how the DBS equipment works. I just assumed it was two electrodes with a constant signal.

[Beartech]

I'm also looking into some of the coordinated reset treatments being used for tinnitus. It is very similar. They have you listen to tones to figure out what frequency your tinnitus is, then they use 4 frequencies (I believe 2 above and 2 below) in a random pattern. I have a friend who has some almost debilitating tinnitus and have pointed him at the research. There is a clinic in London already using this treatment. But is seems really easy to do yourself. I have some tinnitus around 11,000-12,000Hz. I need to look at how the 4 tones relate to the tinnitus frequency.

[danteoh]

@Beartech the study explains the 3:2 is burst time versus rest time. 100ms on, 66.7ms rest. Per finger.

[RChadwick7]

Oh wow! Lots of good info! While I think it's best to stick as close as possible to the professional equipment, spending $300 for each transducer isn't in my budget. I wouldn't be surprised if simple ERM vibrators will do almost the same. I'm going to start with ERM, then also try LRA when they arrive. 250-300Hz shouldn't be to difficult with LRA, piezo, modified speakers or relays. When I have something, I'll share it here, as well as results. If I get good results with ERM, that might make building these things so much cheaper. Does anyone have STL or CAD for fingertip motor housings?

[RChadwick7]

Also, has anyone figured out how the three patterns work? First one looks semi-random, second one looks mostly inverted to the first, and the third is all on. Do we do a few minutes of A, a few minutes of B, then C? Or A, B, A, B, A, B, then finish with C? Something else? I heard these treatments were 3-4 hours, wondering what patterns are happening.

[NeuroMod321]

I'm also intending to start with ERM - as I've been using them for Essential Tremor and HSP for 4 yrs now. The issue with the timing waveform - which has the most unknowns - will need many hours to gauge if it is effective as it will also depend on the severity of the PWP condition. And if the PWP has e.g. Lewey Body Dementia as well - which I assume , will complicate the trial and effects. May take a 100+ hours since no two people react the same to a treatment. Keep up the good team spirit and team effort. Thanks to everyone for contributing and commenting.

[RChadwick7]

While I think we should all help each other when we can, since this is all so experimental it makes sense for everyone to make their own version, then we can copy what works. While I'm happy to share schematics and anything else when I make them up, it doesn't seem to make sense for something unproven and theoretical to be released, 50 people make it, and 50 people realize it doesn't work. If we all shoot in our own direction, someone is bound to succeed, and we could share that with people less skilled. As you say, it might take a LOT of hours before we know it works or not. I'm hoping to have the Electronics done on mine in 3 days or so and write software in a few weeks, if my schedule allows.

[Beartech]

@RChadwick7 the top pattern is vCR, the middle is "noisy vCR" (I believe what the latest study used and what the people seen on the tv story used), and the last is just all on. The Noisy vCR has each cycle actuating each finger pair once in random. The "2-3-4-5" is possible in a truly random system so that's how I read it. I believe it's Noisy vCR, in 3:2 pattern, with every group of 4 being random, for 4 hours. There are 5,040 possible combinations where each group uses all 4 numbers. So in 4 hours there will be some repeats. If it's truly random then some combinations may be used many times over the four hours and some might not show up at all.

[Beartech]

@NeuroMod321 Good ideas. I'll get my three micro speakers I ordered from DigiKey this week so I'm going to start with those. As for software, I will upload some code I wrote last night for the tinnitus acoustic CR I wanted to try. I use mostly Ruby for prototyping and it was very easy to assign 4 tones to variables in an array then use the .shuffle method to randomize the order of the array each time. I just set it to iterate over the newly shuffled array then two cycles of silence over and over.

[Beartech]

I notice on the vCR patterns that each step is .16 seconds and the burst is .1 seconds long. In reading some of the papers I seem to remember something about that .06 pause being of some perceptual importance. Will have to look into that. Will an ER motor be able to spin up to 250hz and back down in the .1 seconds it has to operate? I'm no electrical engineer so I don't know if there is a way (a circuit with a capacitor?) to get the fastest rise out of a motor (both ER and linear)?

[danteoh]

@Beartech yeah the EMR motors I got have a 27ms ramp up and down time.

The 66ms is the estimated time for the neurons to "settle" so that when the next one fires, there isn't this cross-talk effect (is what I understood the purpose).

[NeuroMod321]

@Beartech . Good question. I can think of few options depending on how you are driving the vibrator devices. a) If using a transistor + base resistor , connect a capacitor in parallel with the base resistor. b) Connect a Schottky diode between the base and collector of the driver transistor. c) 3rd. is in part what I am doing for various reasons and testing. Feed a separate PWM over drive V ( vibrator ) voltage ( e.g. to the collector feed of the transistor array ) > ( greater ) than V ( norm ) . E.g. If V( norm ) = 3V , then apply 4V5 to 5V5 to the driver collector lines BUT for a 'wide' period/pulse - i.e. kick start the vibrators. Then reduce the pulse width thereafter. I have used this technique to 'jolt' motors to start them quickly. If you use the ERM coin / pancake motors they start around 2V and even when pulse driven at e.g. 4V5 for prolonged periods - when I check the motors in static state at 2V - they start easily. That is to say - if a motor is driven at >> V( norm ) for long periods , usually there after the V(min ) to start , increases . But those coin motors do not seem to have the "volage memory " effect - which is a good feature - for this application particularly.

[danteoh]

@NeuroMod321 the vybromics Brushless DC ERMa explicitly can't use PWM... as a heads up.

[NeuroMod321]

It works !! - pasting pict. the ERM is encased in the wearable and PWM driven - as per description in previous post.

[danteoh]

Since we're sharing - My box is basically done using said Vybronics ERMs. I can swap finger hardware relatively easily so long as they run on constant 3V for now. Otherwise I'd have to solder up a new protoboard using the PWM pins instead of the dedicated digital pins.

[NeuroMod321]

@danteoh . Fantastic !! great job !! may I ask - why do some fingers have 2 wires and others 4 wires? And how is the person with PWP finding the glove , initially ? - appreciate more time is needed . And the whole module is battery powered ? Very impressive !!

[danteoh]

That's actually just my mother-in-law. It's her sister that we built the device for I was just doing rough fitment of the Velcro.

All fingers have 2 wires, a shared 3v hot and then discrete grounds that go into the transistors. The 4 wire is optical illusion- it's going into her index finger you just don't see it in the pic. I know this for sure because the green ground wire goes into the green index finger which is a different color 3D print.

[danteoh]

So I was playing around with my ERM motors this morning - I had it set up for 100ms vibration 66ms rest time per finger.

I dropped it to 20-40ms vibration and (166 minus whatever my vibration time is) for rest. The intensity drops substantially but the vibration is still felt, which I believe is the intent of the low amplitude tactors. 20ms was extremely weak, 40 ms was "definitely felt a vibration", which checks out because time to peak vibration is in the 25ms range.

[Fenderpicker]

In the paper referenced in the top post, CR RVS is truly random throughout the session.

We apply CR stimulation, where the sequence of stimulation site activation is randomly varied between stimulation cycles of length Ts with equal probability. To exploit the transient desynchronization induced by CR stimuli (Tass, 2003a,b), we use a patterned ON-OFF CR stimulation with n = 3 cycles ON and m = 2 cycles OFF protocol (with cycle duration Ts), where the stimulation is delivered only during the ON cycles, and desynchronization is strongest during the OFF cycles, see (Lysyansky et al., 2011).

Noisy CR RVS includes a jitter of 23.5%. Jitter seems to refer to the onset of the stimulation. So, with regular CR, the onset if the stimulation is at 0ms. With noisy, onset of the stimulation can very from 0ms to 23.5ms.

From the referenced paper:
The noisy vCR pattern was obtained by applying a random jitter to each stimulus onset time s (except for the very first stimulus).

[NeuroMod321]

@danteoh . Hi , Just wondering how is your trial coming ? Any feedback / comments / etc ? Thanks.

[danteoh]

@NeuroMod321 my relative is still waiting to speak to her motion specialist about it. Likely won't start for a month or so.

[NeuroMod321]

@danteoh . Thanks.

[RChadwick7]

I'm finally at the point of coding the timing patterns, and I've got a whole bunch of questions!

The patterns A and B (Regular and noisy) appear to be mirrored to both hands, but the paper mentions possible benefits to not mirroring. Any idea what a non-mirrored pattern would look like?

Can the patterns truly be random? Would there need to be rules, such as no single finger twice in a row?

What I would think of jitter would be a shortening or lengthening of pulses and spaces, but the patterns are different instead.

Would there be any advantages to having random variations in on and off pulse length?

The paper also mentions specific brain frequencies for different PD symptoms. Would or should the frequency of these pulses change based on the patient?

Is it worth implementing the 'C' pattern?

It appears that treatment times are 2 hours. The paper mentions 2 hours, then later that day another 2 hours. Then, the paper mentions 2x2 hours. I'm not sure if 2x2 hours really means 4 hours, or just representing 2 hours, then later another 2 hours.

[Fenderpicker]

I've been trying to put together an initial requirements doc based upon the 2021 paper, along with some videos Dr. Tass has on YouTube. Let me know if anything is missing:

Pattern: 100ms on, 66ms off for each finger (4) per cycle equalling 664ms per cycle

Each finger receives one stimuli per cycle

3 random ON cycles per pattern

2 off cycles per pattern equalling 1.3s

Single vibratory bursts are delivered at periodic times subjected to a jitter that is uniformly distributed within the range of ±23.5% the inter-stimulus intervals.

The first cycle always set to 0.

If jitter was +0, time was reduced to equal 100ms to avoid mutual masking.

A vCR sequence comprises four subsequent vibratory bursts, delivered (on average) within one vCR cycle.

Within one vCR sequence each fingertip (two through five) is activated exactly once.

The CR frequency fCR, i.e., the rate at which the CR sequences are delivered, is 1.5 Hz. Hence, the length of a CR cycle is 667 ms.

Bilateral noisy vCR is administered in a mirrored manner to both hands, coincidently activating right and left fingers (two through five).

Vibration frequency is 250 Hz and duration of vibration bursts is 100 ms.

Peak vibration amplitude is .06 to .1mm

The duration of a single vCR session is 2 h. Two sessions a day. One in the morning and one in the afternoon/evening.

Thoughts?

[danteoh]

We're a few days into my aunt trying out the gloves. I'll say results have been promising so far - we're conducting a few motor activities that we measured before the first session, and day-over-day her overall speed to complete activities has been increasing.

[NeuroMod321]

Hi @danteoh - that sounds promising. Does she have any tremors ? And her writing ? Can she write " The quick brown fox jumps right over the lazy dog " with all letters the same size and legible ? Can she draw a clock wise and anti clockwise normally or with serrations ? Sorry for so many questions but this update is interesting. Thanks.

[danteoh]

@NeuroMod321 her legibility has improved quite a bit... less dragging of the pencil. Tremors seem reduced but still present. The most noticeable change has been her walking/changing direction.

[NeuroMod321]

@danteoh . Thanks for feedback. Does she have freezing of gait ? and can she turn around 360 degrees ? Is she using a walker ? Is the controller dis-connected with those observations or off ? Can she move with the controller ON ?

[danteoh]

Historically, she does have freezing and shuffles to turn around/do a 360, and uses a walker. She's had very little freezing the past 24 hours in particular and has been able to turn while walking without dragging and shuffling her feet.

She still uses a walker throughout the day sometimes but she's said she feels pretty confident without it.