Products
The array is the brain. The add-ons are the body.
Everything we make has two parts. The HemiSense Array is a wearable sensor network that reads the nervous system, understands what it reads, and explains itself. On its own it can detect, monitor and predict. HemiAct add-ons plug into the array and act on what it knows. SureHand, a tremor cancellation sleeve beginning with Parkinson's disease, is the first of them.
The HemiSense Array
A wearable network of eleven sensors at seven sites on the body, in development. It reads signals from the brain, the nerves, the muscles and the heart, and it turns them into something a person can act on and a doctor can read. It is a product in its own right. Worn alone, with no add-on attached, it does three things.
Detect
Find the signature of a neurological condition in daily life, including early signs that a twice-yearly visit would miss. Ask the wearer to open and close their hands, read the signal, and know what it means.
Monitor
Follow a condition over weeks and months on a continuous scale, so a clinician sees the whole record rather than a description from memory.
Predict
Model where things are heading and how soon, with the confidence and the reasons stated in plain rules rather than implied by a score.
Seven sites. Three on the midline, four worn on both sides. Eleven sensors in all.
- 1 Headband midline
- 2 Elbow, both arms building first
- 3 Wrist, both arms building first
- 4 Chest strap midline
- 5 Waist belt midline
- 6 Knee, both legs
- 7 Ankle, both legs
Modular sensors for personalized care. A wearer puts on only the sites their care needs. A person with tremor needs the arm. A person with a gait problem would wear the waist, knees and ankles. A person with seizures would wear the headband. The software and the way it explains itself stay the same across all of them.
What we are building first. The elbow and wrist sensors, and the SureHand sleeve that runs on them. The other five sites come after, one at a time, as the evidence allows.
HemiAct add-ons
The array knows what is happening. A HemiAct add-on is anything that does something about it. Each one plugs into the array, borrows its senses and its judgment, and carries out an action: a signal to a muscle, a dose of a medicine, a command to a machine. The add-on carries the fourth function, Treat, and every add-on inherits the same rule: what it does can be read and checked by a doctor.
First, in development
SureHand sleeves
A Parkinson's tremor cancellation sleeve for the forearm. It opposes tremor with a small signal timed against it, using the elbow and wrist sites. The full description is below.
Future, exploratory
Adaptive dosing
An add-on that gives a medicine in the amount the body needs at that moment, read from the array, so a person is neither over nor under medicated across a day. This is an idea we are exploring. It is not in development.
Future, exploratory
Interfaces to other devices
The array can speak to machines other people build. We see a path to controlling a computer, a powered wheelchair, a drop foot device, or a robotic prosthetic hand from the signals the array already reads.
SureHand sleeves
A forearm sleeve, in development, for people whose hands shake. It reads the signal your nervous system sends to your hand, predicts the tremor riding on that signal, and answers with a small signal of its own, timed to push against the shake so the movement you meant comes through.
It is soft, it is worn like a compression sleeve, and it has no wires running to anything else. The two round sensors are the HemiSense Array at the elbow and the wrist. The sleeve between them is the HemiAct add-on: elastic fabric with conductive thread woven through it, which is what delivers the counter-signal.
Design rendering. Sensor at the elbow, sensor at the wrist, woven thread between. Not yet a manufactured prototype.
The loop, four steps
01
Read at the elbow
The array's elbow sensor picks up the motor signal on its way to the hand, before the hand has moved.
02
Separate and predict
The array splits the movement you intended from the tremor riding on it, and predicts the phase the tremor will be in next.
03
Push back
Fine conductive wire woven into the elastic fabric delivers a small impulse across the forearm, timed out of phase with the tremor, so the tremor is opposed and the intended signal passes through. This is the add-on's job.
04
Check and tune
The array's wrist sensor measures what the hand actually did and feeds that back, so the model keeps tuning itself to this wearer, this day, this hour.
The SureHand loop. Two array sites, one add-on, running continuously. The sensors and the software belong to the array. The impulse belongs to the sleeve.
The research underneath
Each step of the loop stands on published work.
For every step, here is what prior studies have shown, and the claim SureHand makes on top of them.
Dosen and colleagues showed that tremor can be detected online from the electrical activity of the very muscles that produce it, early enough to counter it in real time.[1]
So read it there, at the source. The muscles that move the hand lie in the forearm, and their signal can be picked up near the elbow. Reading upstream of the wrist is how we aim to buy the controller its lead time.
Prochazka's group showed in 1992 design work that a closed-loop controller could suppress the 2 to 5 Hz rhythm of pathological tremor while barely touching voluntary movement below 1 Hz.[2] Grossman's group later showed that timing decides everything: stimulation locked to the tremor's phase reduced it, and the same stimulation without phase locking had no significant effect.[3]
So the hard part is telling intent from tremor and timing the next beat, and both must be readable. We do each with fuzzy rules a clinician can check, because a timing guess nobody can inspect is a dose nobody can check.
With out-of-phase stimulation of the opposing muscle, tremor fell in five of six patients: by 46 to 81% at motor-level intensity and by 35 to 48% at gentler, sensory-level intensity.[1] The same cancellation principle reduced resting tremor by almost half over the motor cortex,[4] and by up to 87% in the thalamus, in selected patients, using less than half the energy of conventional deep brain stimulation.[5]
Phase cancellation has now worked at the cortex, the thalamus and the muscle. What it has lacked is a form someone can wear all day. SureHand is designed to put the counter-signal into the fabric at the forearm, with no surgery, no headset and no box on a belt.
A home study of scheduled wrist stimulation (263 patients enrolled, 205 completed) found accelerometer-measured improvement in 92% of the patients measured.[6] Yet reviews of the field report widely varying methods and mostly small samples,[7] and warn that tremor's natural swings in intensity can blur results.[8]
Tremor is not constant, so the therapy cannot be either. The wrist sensor is designed to grade every response as it happens and retune the model to this wearer, this day, this hour. Measurement is part of the treatment, not a step after it.
Rudin argues that when the stakes are high, models should be interpretable by design rather than explained after the fact.[9]
A device that sends current into a person's arm is exactly that kind of decision. That is why the engine is fuzzy logic: the rules are the model, and a doctor can read them.
These studies test related methods, not SureHand. SureHand has not yet been tested in patients.
References (9)
- Dosen S, Muceli S, Dideriksen JL, Romero JP, Rocon E, Pons J, Farina D. Online tremor suppression using electromyography and low-level electrical stimulation. IEEE Trans Neural Syst Rehabil Eng. 2015;23(3):385-395. doi:10.1109/TNSRE.2014.2328296
- Prochazka A, Elek J, Javidan M. Attenuation of pathological tremors by functional electrical stimulation I: Method. Ann Biomed Eng. 1992;20:205-224. doi:10.1007/BF02368521
- Schreglmann SR, et al. Non-invasive suppression of essential tremor via phase-locked disruption of its temporal coherence. Nat Commun. 2021;12:363. doi:10.1038/s41467-020-20581-7
- Brittain JS, Probert-Smith P, Aziz TZ, Brown P. Tremor suppression by rhythmic transcranial current stimulation. Curr Biol. 2013;23(5):436-440. doi:10.1016/j.cub.2013.01.068
- Cagnan H, et al. Stimulating at the right time: phase-specific deep brain stimulation. Brain. 2017;140(1):132-145. doi:10.1093/brain/aww286
- Isaacson SH, Peckham E, Tse W, et al. Prospective home-use study on non-invasive neuromodulation therapy for essential tremor. Tremor Other Hyperkinet Mov. 2020;10:29. doi:10.5334/tohm.59
- Pascual-Valdunciel A, Hoo GW, Avrillon S, Barroso FO, Goldman JG, Hernandez-Pavon JC, Pons JL. Peripheral electrical stimulation to reduce pathological tremor: a review. J NeuroEng Rehabil. 2021;18:33. doi:10.1186/s12984-021-00811-9
- López-Blanco R, Benito-León J, Hernández-Gallego J, Sánchez-Ferro Á. The validation of tremor-cancelling technologies needs a multidisciplinary consensus statement. Tremor Other Hyperkinet Mov. 2020;10:530. doi:10.5334/tohm.530
- Rudin C. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat Mach Intell. 2019;1:206-215. doi:10.1038/s42256-019-0048-x
What it is made of
An elastic sleeve worn on the forearm. Conductive thread is woven into the fabric itself, so the sleeve carries the impulse and there is no rigid housing strapped to a moving arm. The elastic holds the thread against the skin as the arm bends and turns, and the open weave lets the skin breathe under it for a full day. Two array sensors, one at the elbow and one at the wrist, each a flat disc that sits in the fabric. Everything between them is fabric.
It is designed to be put on in the morning, without help, with one hand if the other is shaking, and then left alone.
Where it stands. SureHand is in development. It has not been tested in patients. It is not for sale, and it is not cleared or approved by any regulator. We are working toward a first feasibility study with clinical partners in movement disorders. When that begins, this page will say so.
How it grows
More senses, then more hands.
The array grows one site at a time. Each new site adds something the brain of the system can read: the headband for seizures and cognition, the waist and legs for gait. The add-ons grow alongside it. Each new HemiAct add-on takes what the array already knows and puts it to a new use.
We do not start over for each condition. We add a sense, we write the rules for it, and we build the hand that acts on it. Everything underneath stays the same, including the promise that it can explain itself.
In plain terms. Brain first, then body. The HemiSense Array is the product we are building the company on. SureHand is the first thing it does with its hands.
Want the engineering underneath this?
The Technology page has the sensing modalities, the inference engine, the research it rests on, and the validation path.