DK201700437A1

DK201700437A1 - A Novel Cyber-Organic Motor-Neural Interface

Info

Publication number: DK201700437A1
Application number: DKPA201700437A
Authority: DK
Inventors: Bulow Pedersen Emil
Original assignee: Talos Cybernetics ApS
Priority date: 2017-08-07
Filing date: 2017-08-07
Publication date: 2019-02-25
Also published as: DK181212B1; WO2019029777A1

Abstract

The invention comprises a novel cyber-organic motor-neural brain-computer interface for the purpose of extracting intended and or actual movement from a human or animal. The interface converts motor-neural signals to intended contraction data of the recipient muscle. This is accomplished using a pipeline of organic, mechanic and electronic components each responsible for translating the neural signals to intermediate representations that are eventually converted to a digital or analog signal describing the contraction value encoded in the original neural signal. The command to contract a muscle begins in the central nervous system 1. It is transmitted through multiple motor neurons 2 and their axons 3, each innervating multiple muscle fibers of the organic component 4. Each fiber contracts upon receiving an action potential, resulting in a mechanical signal. This signal is transmitted to the compliant mechanical component 5, which converts the mechanical signal to an analog electrical signal of proportional intensity.

Description

Field of the invention

The invention relates to the field of brain-computer interfaces - that is, extracting intention from the nervous system of a human or animal. It is targeted towards the collection of motor data from the human nervous system, it aims to solve the major problems of the low resolution, low bandwidth neural interfaces that are currently used in bionics, neuroprosthetics and gesture control devices.

Prior Art

The invention comprises a novel motor-neural brain-computer interface for extracting intended and or actual movement from a human or animal. The interface converts motor-neural signals to intended contraction data of the recipient muscle. This is accomplished using a pipeline of organic, mechanic and electronic components each responsible for translating the neural signals to intermediate representations that are eventually converted to a digital or analog signal describing the contraction value encoded in the original neural signal.

Known methods of extracting intended movement data from the motor-neural system include but are not limited to:

• Electromyography • Electrocorticography • Electroencephalography • Myooximetry • Electrogoniometers • Switches, accelerometers, pressure sensors and other traditional controls

DK 2017 00437 A1

Electromyography (EMG), which measures the average action potential over an area of axons or muscle tissue, works by letting the electrical field caused by the polarization changes during action potential in the axons of motor units induce a voltage between two electrodes placed in close proximity to the measured tissue. This voltage is then amplified, rectified, and filtered. Even at this point, the signal mostly looks like noise. Generally, the amount of noise is seen as an indicator of intended contraction strength. Not only is this not a very precise measure of intended contraction - the signal varies greatly between muscle activation, sustained contraction and with fatigue, making it even harder to accurately predict the intended contraction value of a particular muscle. Furthermore, surface EMG requires direct skin contact, whereas deep EMG requires electrodes to be directly exposed to the measured tissue. This last connection can degrade considerably over time due to the formation of scar tissue around the electrodes of the system.

It is also known, that the noise caused by interference of the electrical fields of action potentials from different motor units in the same nerve bundle can be somewhat mitigated by growing the axons into a grid of channel-like electrodes. This allows the signal to be based on the temporal and spatial integration of action potentials, as opposed to the average of all nearby potentials, where potentials from a large area interfere with each other.

It is also known, that in situations where the muscle tissue once connected to a motor nerve has been removed, the strength of the electrical fields produced during intended contraction can be amplified considerably by surgically attaching and innervating muscle tissue at the end of the motor nerve bundle.

Electrocorticography (ECG) fundamentally works in the same way as EMG. The main difference is that the electrodes are placed directly onto the surface of the brain, and that they are usually packed into arrays of many ECG channels. The outputs of these channels are combined using machine learning algorithms that are trained to recognize

DK 2017 00437 A1 the activation patterns associated with different movements. While this works relatively well, it is highly invasive and requires a long period of training prior to efficient use.

Electroencephalography (EEG) works by electrically detecting neural activity noninvasively through the scalp in a fashion similar to EMG. It contains many electrodes combined using the same types of algorithms and training used for ECG. One major difference is that EEG electrodes are placed all over the scalp, meaning that they get signals from many different areas of the brain. The result is that abstract brain centers are suddenly responsible for performing movement related tasks in addition to their original tasks. This, coupled with the poor signal quality from measuring through the scalp, results in very poor bandwidth and a high need for attention from the user.

Myooximetry works by detecting the amount of oxygenated blood within a muscle. During contraction, this amount increases to meet the increased oxygen and energy demand of the muscle. This is also not a very accurate measure of muscle contraction, since it suffers from many of the same problems as EMG. Furthermore, the delay from activation to increased blood flow is non-negligible.

Electrogoniometers are essentially protractors for skeletal joints. They require the entire joint to be present, and furthermore requires the placing of sensing elements on both members of the joint. This makes it next to useless for extracting intention in bionics, neuroprosthetics and consumer products where required limbs and tissue may be unavailable for various reasons.

DK 2017 00437 A1

Switches, accelerometers, pressure sensors and other traditional controls have traditionally been used as a primitive way to control simple prostheses. Recently, through the work of DEKA, more advanced control schemes have been developed, yet a common problem persists: A separate limb is needed to operate these types of controls. This is often seen in the shape of a pressure sensor in the shoe or an accelerometer on the shoulder. Controlling the bionic limb thus interferes with the normal operation of the controlling limb, making it unfavorable for seamless integration with the nervous system.

The present invention eliminates most of these problems:

• It has very high contraction resolution • Very low delay from neural command to decoded digital signal • Leaves no electrodes exposed to living tissue, meaning all electronic components can be coated in biocompatible materials.

• Immune to changes in signal over time due to fatigue, scar tissue, sweat and other traditionally challenging obstacles.

• Integrates action potentials temporally from many motor units to produce an accurate, absolute intended contraction value for any given muscle - whether or not the original recipient muscle is intact or even present.

This makes the invention a prime candidate for use in controlling bionic limbs. Here it can allow direct 1:1 mapping of intended contraction of each muscle in a given amputated limb to actual movement in a bionic replacement limb. The high resolution and low latency are the key elements of the invention that sets it apart from the known art. Previously a move had to be planned, transmitted to the bionic limb through a series of muscle contractions and *then* executed before the result of the movement could be seen and reacted upon. With this new technology, the feedback

DK 2017 00437 A1 loop between the body's senses and the movement of the bionic limb can be made small enough for the patient's motor cortex to perform real time corrections across a large number of cooperating muscles. This allows movement to be faster, more accurate and completely natural. Since the heavy tasks of planning and executing complex movement patterns is offloaded to the patient's own motor cortex, the required attention from the patient will be the same as for a biological limb. This means the bionic limb will feel almost exactly like controlling its biological counterpart. The inclusion of the motor cortex can be likened to the difference between rendering 3D graphics on a CPU versus rendering on a GPU - that is, the motor cortex is a highly specialized center capable of performing complex movements at a far lower attention cost than other parts of the brain. This invention allows bionics to tap into that power.

Furthermore, in the field of neuroprosthetics, it can help bypass damaged nerve tissue by determining the intent of the patient, in order to artificially stimulate neurally disconnected muscle tissue accordingly. One example of such a situation is drop foot, where the peroneal nerve is damaged, making it difficult for the patient to raise their foot when walking. Control systems for artificially stimulated dorsiflexion have so far been accomplished using a heel switch and a timer, resulting in a rigid gait incapable of adapting to advanced types of terrain.

Overview and Flow of Information

The interface converts motor-neural signals to electrical analog or digital signals representing desired muscle contraction levels encoded in the neural signal. It does so by using muscle tissue as an intermediary translator, allowing the desired contraction level to be decoded as mechanical deformation.

DK 2017 00437 A1

The flow of information looks as such:

1. A move is planned in the central nervous system.

2. It is then transmitted to muscle tissue via the somatic nervous system as neural commands describing the desired contraction level.

3. Muscle tissue contracts according to the neural signal, effectively transforming it to a mechanical signal.

4. The mechanical signal is transformed to an electrical signal by a muscle deformation sensing component.

5. The electrical signal is filtered, processed and if needed, converted to a digital signal, which describes the contraction level encoded in the neural signal.

6. The analog or digital signal is used in an application. This may include combining multiple interface instances, in order to extract more complex intended actions.

Points 3, 4, and 5 combined form a direct cyber-organic motor-neural interface. When modeled as such, a neural signal enters one end of the interface, and an analog or digital signal describing the decoded level of contraction exits the other. Point 3 is referred to as The organic component”, point 4 is referred to as The mechanical component, and point 5 is referred to as “The signal processing component”.

Organic Component

In many cases, the user’s existing unmodified skeletal muscles are sufficient for use as the organic component. In these cases, the mechanical component can optionally be connected non-invasively, by letting the skin act as a mechanical transmitter between the organic and mechanical components. Due to the imperfect mechanical transmission of the skin coupled with the fact that it may receive mechanical signals from multiple sources, this mode of operation often sees a fair amount of motion

DK 2017 00437 A1 artifacts present in the output signal. These are often caused by mechanical interference from nearby muscles but can also be caused by nearby joints. This can be mitigated by combining multiple interface instances using machine learning or sensor fusion algorithms. It can even be combined with other types of neural interfaces like e.g. electromyography, In this case, EMG would be used to correlate the deformation of muscle tissue with the activation of the nerve that innervates it, in order to verify that the detected deformation did in fact originate in the measured tissue.

In cases where no sufficient muscle tissue is available at the end of a given motornerve, a replacement is needed. This can either be acquired by routing the nerve to nearby healthy muscle tissue, grafting muscle tissue from elsewhere in the body, or by growing muscle tissue from scratch. This tissue is surgically attached to the nerve using targeted muscle reinnervation (TMR).

Mechanical Component

The mechanical component is affixed on, near or within the muscle tissue of the organic component. It transforms the mechanical signals generated by contracting muscle tissue into an electrical signal, which can be further processed for use in an application.

The sensing component relies on the fact that the stiffness and diameter of a skeletal muscle changes as the muscle contracts. As such, the spatial deformation between a series of points in a muscle is proportional to its contraction level.

The mechanical component in its most basic form consists of a deformable member onto which a photo-emitter and a photo-detector are placed in such a way that the

DK 2017 00437 A1 light hits the photo-detector at maximum intensity, when the deformable member is in its resting position. The resting position is defined as the state in which the deformable member experiences the (east degree of deformation while being fully mechanically compliant with the organic component onto which it is placed.

When the organic component deforms, the deformable member deforms with it, decreasing the amount of light that hits the photo-detector, as shown in drawing 2.

In another embodiment part of the deformable member is stiffened as shown in drawing 3. The stiff part casts a shadow onto the photo-emitter, creating a much sharper divide between light and dark. This results in the mechanical component being much more sensitive to minute deformations, making it more suited for detecting the tiny deformations of some embodiments of the organic component. This embodiment is particularly well suited to manufacturing using flexible printed circuits (FPCs). In an example hereof, the deformable member is the flexible circuit itself with appropriate stiffening, the photo-emitter is a side mounted SMD LED, and the photo-detector is a side-mounted SMD photoresistor, phototransistor or reverse biased LED,

In another embodiment one or more opaque blockades are placed along the deformable member as shown in drawing 4. The blockade casts a shadow onto the photo-emitter, creating a much sharper divide between light and dark. This results in the mechanical component being much more sensitive to minute deformations, making it more suited for detecting the tiny deformations of some embodiments of the organic component. This embodiment is especially sensitive to deformations that occur close to the blockade(s) along the deformable member.

Signal Processing Component

It is often beneficial to normalize the signal, in order to abstract away some of the physical characteristics of the organic component, allowing a contraction value to be

DK 2017 00437 A1 seen as a fraction of the maximum contraction value with respect to the minimum. These extrema can be found through calibration.

Since many applications are interested in the intended degree of contraction, linearization can be applied in cases where the degree of deformation does not map linearly to the intentions of the user.

Furthermore, every component in the pipeline introduces a certain amount of noise into the signal. This noise can be filtered out at this stage.

Extraction of Complex Motor Data

Many types of movement are produced by several muscles working together in complex patterns. This means that knowing the intended contraction values of individual muscles will not necessarily allow the intended joint geometry to be known. This can be solved by using multiple sensor instances to get data from several muscles. A machine learning model can then be trained to estimate joint geometry from muscle contractions using both intended contractions and actual movement data acquired from both healthy individuals and amputees.

Applications

Neurorehabilitation

The present invention can help greatly with regaining proper movement after neuropathy or brain damage. By amplifying the feedback received from the partial contraction of muscles whose nerves have been damaged, the patient can more effectively identify actions that result in contraction, helping them regain control of muscles and stimulating a higher rate of nerve regrowth. Furthermore due to the high bandwidth and digital nature of the interface, it can be used as input for computer

DK 2017 00437 A1 games, allowing patients to have fun while performing otherwise very monotonous and repetitive tasks. This has been shown to increase patient retention rates and speed up recovery. Furthermore, the mental wellbeing of patients is known to be an important aspect of the recovery process.

In addition to the direct benefits to patients, the interface provides valuable data for doctors and therapists to track progress.

Physiotherapy

The present invention can help greatly with rehabilitating muscles after injuries or medical operations. By using the digital signals produced by contraction of the muscles that are to be rehabilitated as input for a computer game or simulation, exercises can be controlled and monitored accurately by therapists. Furthermore this allows patients to have fun while performing otherwise very monotonous and repetitive tasks. This has been shown to increase patient retention rates and speed up recovery. Furthermore, the mental wellbeing of patients is known to be an important aspect of the recovery process.

Perineometry

Known perineometers are either very impractical and obstruct use of the legs or require insertion into either the vaginal or anal cavities. A perineometer can be constructed by configuring the invention to use the perineum as the organic component and optimizing the mechanical component to its long muscles and minute deformations during contraction. This allows a far more comfortable and non-invasive approach to treatment of urinary incontinence for both male and female patients.

DK 2017 00437 A1

Research of neurodegenerative diseases

A great problem in the research of neurodegenerative diseases such as Parkinson’s and others is that researchers can only gather movement data in a laboratory setting. Therefore, the data gathered is not necessarily an accurate image of the patient's conditions, as symptoms often vary over time and between situations. Attempts to overcome this have previously been made using portable EMG dataloggers and motion processing units located in various regions of interest across the body.

Due to the aforementioned disadvantages of EMG and the fact that the motion processing units only measure joint geometry; this solution provides incomplete data at best.

In this situation, the present invention is able to log the exact contraction value of tremors at high sample rates and high resolution. In addition to more accurate data collection over time, this will also allow researchers to study and compare the waveforms of individual contractions. This will allow researchers to more accurately study neurodegenerative diseases as well as study the effect of various drugs on the nervous system of the patient.

Neuroprosthetics

A great problem in neuromuscular prosthetics is determining the intention of the user. Luckily most joints are actuated by more than one muscle, so even if one becomes impaired, its movement can often be approximated by looking at the contraction values of one or more of the other muscles.

An example of how this would work can be seen with drop foot. In patients with drop foot the tibialis anterior is paralyzed, leaving the patient unable to lift their foot,

DK 2017 00437 A1 resulting in an abnormal, often unhealthy gait, in this situation, the contraction levels of the calf muscles can be used to predict the patient's intent and estimate the desired contraction value of tibialis anterior, which can in turn be stimulated accordingly.

Another problem in neuro prosthetics is accurate stimulation of muscles. Traditionally motion processing units have been used to establish a closed feedback loop between stimulator and the stimulated limb. This feedback loop has room for drastical improvement however, since joint geometry is a very indirect and delayed reaction to muscle contraction.

By configuring the neural interface to use the muscle to be stimulated as the organic component, the contraction level resulting from the neural commands sent by the artificial muscle stimulator can be known with very little delay. This level can in turn be compared to the desired contraction level, and the stimulation signal can be adjusted accordingly. This allows any contraction level to be induced and held, and even allows the desired contraction level to change over time, resulting in smooth continuous artificially stimulated movement.

Bionics

One of the most important aspects of bionics is allowing the patient to intuitively control the bionic limb. This means that the bionic limb must be able to extract or estimate the intentions of the patient. As previously described, most presently known methods of control provide low bandwidth, slow operation and require a significant amount of attention from the patient.

Since the present invention allows decoding of intended contraction values for muscles with very high resolution and low delay, it is a prime candidate for the job. Some types of amputation, such as transradial and transtibial amputations allow most of the

DK 2017 00437 A1 muscles controlling the amputated limb to persist in the stump post amputation. In these cases, the neural interface can be configured to use the remaining muscle tissue as the organic component. In the non-invasive embodiments, this allows the creation of inexpensive, mass produced bionic limbs. Alternatively, the interface can be 5 implanted in order to have perfect placement of the interface every time without effort from the patient.

In cases where more degrees of freedom are desired, or where the type of amputation doesn't leave as much usable muscle tissue, the neural interface must be configured to use surgically attached muscle tissue as the organic component. By combining multiple 10 interface instances, this allows 1:1 mapping of the desired contraction of each muscle in the amputated limb to actual movement in a bionic limb.

DK 2017 00437 A1

Descriptions of drawings

Drawing 1

Drawing 1 illustrates the pipeline that decodes motor neural signals, outputting a digital or analog value describing the contraction level encoded in the neural signal.

The command to contract the muscle 4 begins in the central nervous system 1, It is transmitted through multiple motor neurons 2 and their axons 3, each innervating multiple muscle fibers of the organic component 4. When each muscle fiber receives an action potential it contracts, resulting in mechanical deformation of the organic component. This mechanical signal is transmitted to the compliant mechanical component 5, which converts the mechanical signal to an analog electrical signal of proportional intensity. This signal is optionally converted to a digital signal and optionally further processed by a computer 6. The resulting digital signal is an absolute representation of the intended contraction level originally encoded in the signal sent by the central nervous system.

Drawing 2

Drawing 2 illustrates the basic construction of the mechanical component.

Drawing 2.a is a side view in the resting position

Drawing 2.b is a side view in a deformed position

Drawing 2.c is a top down view in the resting position

The mechanical component consists of a deformable member 1 onto which a photoemitter 2 is placed such that the primary direction of the emitted light points directly towards a photo-detector 3, which is pointed directly towards the emitter when the deformable member 1 is in its default resting position.

DK 2017 00437 A1

When the member is in the resting position as in drawing 2.a and drawing 2.b, the Light from 2 hits 3 at maximum intensity. When the member deforms as in drawing 2.c, the fraction of photons from the photo-emitter 2, that successfully enter the photodetector 3 is decreased in proportion to the degree of deformation of the deformable member 1.

Drawing 3

Drawing 3 illustrates an embodiment of the mechanical component from drawing 2 which is more sensitive to minute deformations.

In this embodiment parts 4 of the deformable member 1 have been stiffened, such that they do not deform as easily as the rest of the member 1. The result is that the shadows of the stiffened parts 4 have their penumbras projected upon the photodetector 3, such that there is a much sharper transition from light to dark. This allows the output signal to be much more sensitive, meaning that it can be calibrated such that even tiny deformations completely saturate the sensing component.

The stiffened parts 4 do not need to be present in both sides, and their length can be varied to vary the active mechanical band of the component, in order to calibrate it to different types and sizes of the organic component.

Drawing 4

Drawing 4 illustrates an embodiment of the mechanical component from drawing 2 which is more sensitive to minute deformations.

DK 2017 00437 A1

In this embodiment one or more opaque blockades 4 are placed onto the deformable member 1, such that its movement is perpendicular to the direction of the light emitted from the light emitter 2, The blockade casts a shadow onto the photodetector in a way such that the position and width of the penumbra can be controlled by varying the height and position of the blockade, respectively. Upon deformation of the deformable member 1 the blockade(s) 4 moves relative to the photo-emitter 2 and photo-detector 3, such that the intensity of light detected by 3 is proportional to the deformation of the deformable member 1.

Drawing 5

Drawing 5 illustrates the orientation of the mechanical component relative to the muscle fibers of the organic component. It shows several different scenarios, all of which are valid orientations.

The mechanical component 2 is placed on, in or near the organic component 1, is such a way that mechanical deformation of the organic component 1 will be transferred to the mechanical component 2. The drawing serves to illustrate that any orientation is valid, as long as the organic component 1 experiences adequate mechanical deformation in the plane of deformation of the mechanical component 2.

Drawing 6

Drawing 6 illustrates an embodiment wherein the organic component is one of the patient’s existing skeletal muscles.

DK 2017 00437 A1

In this embodiment, the mechanical component 2 is placed directly above the desired muscle in the limb 1, in such a way that contraction of the muscle results in deformation of the mechanical component 2. The output of the mechanical component is fed to a computer 3, which can optionally digitalize the analog signal and or perform signal processing.

Drawing 7

Drawing 7 illustrates an embodiment wherein the organic component is one of the patient s existing skeletal muscles, and wherein the mechanical component is surgically implanted.

In this embodiment, the mechanical component 2 is implanted near or in the desired muscle in the limb 1, in such a way that contraction of the muscle results in deformation of the mechanical component 2. The output of the mechanical component 2 is fed to an implanted computer 3, containing a means of wireless power reception and a transmitter for communicating the digital representation of the original neural signal to an external receiver not shown in the drawing. The computer 3 receives power wirelessly from an external wireless power transmitter located in the near vicinity of the power receptor of the computer 3.

Drawing 8

Drawing 8 illustrates an embodiment wherein the nerve whose signal is to be decoded has been damaged or transected. This means that no existing skeletal muscle tissue can be used as the organic component without prior modification.

In this embodiment, the damaged nerve 2 in the limb 1 (shown here as the stump of an amputated limb) is surgically attached to the organic component 3 at the transection

DK 2017 00437 A1 site 5 using targeted muscle reinnervation (TMR). The muscle tissue that makes up the organic component 3 can come from a variety of sources, including but not limited to muscle tissue grown from scratch, grafted from elsewhere in the body, or tissue still located elsewhere in the body, in which case the nerve 2 must be rerouted to reach said tissue. The mechanical component 4 is attached to the surface of the organic component 3 either by attaching the ends of the elongated member of the mechanical component 4 to the tendons at the ends of the organic component 3, or by attaching the mechanical component 4 to an elastic biocompatible pouch encapsulating the organic component 3.

Drawing 9

Drawing 9 illustrates extraction of complex motor data by combining multiple interfaces.

Several instances 1 of the neural interface are connected to the nerves of a limb 2. The contraction values reported by each instance 1 of the neural interface are collected in a computer 3. This computer then applies machine learning algorithms to extract the desired joint geometry from the desired contraction values of the muscles that would normally produce it. This works no matter which embodiment is used for the neural interface, and whether or not the actual joints whose intended geometry is estimated actually exist.

Drawing 10

Drawing 10 illustrates an embodiment that allows precise closed loop contraction control during stimulation of muscles.

DK 2017 00437 A1

An artificial muscle stimulator 1 stimulates the nerve 3 of a muscle 4, which acts as the organic component of the neural interface. The exact method of stimulation is unimportant, as long as the intensity of the stimulation can be modulated in real time. The resulting contraction is transmitted to the mechanical component 5, which converts the contraction to an electrical signal, which is transmitted back to the artificial muscle stimulator 1. The actual contraction level is then compared to the desired contraction level by 1, and the stimulation signal is modulated according to the algorithm described in drawing 11.

io Drawing 11

Drawing 11 is a diagram illustrating the algorithm used to modulate the stimulation signal of the embodiment described in drawing 10. It is the same algorithm as used in servo motors, and it is shown here in its simplest form.

Claims

Claims

1. A system for extraction of complex motor data from the nervous system of an animal or human comprising:

• An organic component that translates a motor-neural signal encoding a degree of muscle contraction to a mechanical signal of proportional intensity.

• A mechanical component responsible for translating the mechanical signal to an electrical signal.

• Optionally a signal processing component responsible for converting the electrical signal to a digital signal and/or processing the signal.
2. The system described in claim 1, wherein the organic component consists of existing muscle tissue of the organism whose motor-neural activity is being measured.
3. The system described in claim 1, wherein the organic component consists of muscle tissue that is surgically attached at the end or transection of a bundle of motor-neural axons or grown using cells from the patient.
4. The system according to claim 2 or 3 wherein the mechanical component measures the cumulative deformation caused by the contraction of a series of muscle fibers connected to the axons of some or every motor unit responsible for innervating a certain muscle.
5. The system according to claim 4 wherein the mechanical component conforms to the deformation of the organic component, and in which the deformation of the mechanical component results in a proportional change in the intensity of its electrical output signal.

DK 2017 00437 A1
6. The system according to claim 5 wherein the mechanical component contains an optical emitter and an optical sensing element affixed to an elongated deformable member in such a way that the drop in light intensity between the optical emitter and optical sensor, and thus the electrical output signal of the mechanical component, is proportional to the degree of deformation of the elongated member and thus the organic component.
7. The system according to claim 6 wherein the light path along the elongated member of the mechanical component contains light absorbing partial blockades that control the active band of the mechanical component allowing the mechanical component to closely adjust to match the properties of the organic component chosen for a specific application, thereby increasing the effective resolution of the interface.
8. The system according to claim 7 wherein the light absorbing blockades are implemented by stiffening parts of the elongated member, such that they will partially or fully block the light path upon deformation.
9. The system according to any of the previous claims wherein the organic component is the patient's existing muscle tissue, but wherein the origin of the neural commands is an artificial muscle stimulator, such that the effect of the stimulation can be seen in real time, such that the stimulator can perform real time corrections in order to reach a desired contraction level.
10. The system according to any of the previous claims wherein multiple interface instances are combined using machine learning algorithms such that complex joint geometry and force data can be extracted from the contraction values reported by individual interface instances.