Natural movements with a prosthetic hand

Prosthetics specialist Ottobock is researching a new method that will allow arm prosthetics to be moved more intuitively and fluidly at its Vienna site. The prosthesis learns to recognize patterns in the muscle signals and to translate them into motion. It adapts to the wearer, instead of the wearer adapting to the prosthesis.

People who have lost an arm in an accident can often receive a relatively good replacement for their limb. So-called myoelectric prostheses recognize the activity of remaining muscles and turn these into movements with electric motors. This allows artificial hands to be used well with enough practice. The axes of movement can only be used individually, and the wearer must switch between them. In the future, it is to become easier to control arm prostheses. The prosthesis is to adapt to the wearer, learn from his or her muscle signals and translate them into fluid motions.

Pattern recognition is the key that is to eliminate operations and long adjustment periods for prosthesis users and make life easier for them. At the technology center of the healthcare technology company Ottobock in Vienna, specialists are working on preparing this pattern recognition for series production and on optimizing the associated hardware and software.

To understand exactly what pattern recognition improves, you first have to look at what kinds of prostheses are used for arm amputees today.

Prostheses today
The first is the cosmetic prosthesis. It cannot do much aside from replacing the lost arm in terms of its shape and weight. For simple movements, a person at least needs a cable-activated prosthesis. With this type of prosthesis, movements can be completed mechanically with the existing muscles and cable mechanisms.

The next step up is the myoelectric prosthesis. With hand amputees, two sensors on the top and bottom of the forearm measure muscle signals and turn them into movements using small electric motors. This allows an artificial hand to be opened and closed, for example. To rotate the hand, multiple muscles can be tensed at the same time to switch the axis of movement.

This makes it possible to rotate the hand, move it up and down, grip an object with the tips of the thumb and forefinger (pincer grip) or to hold a key (key grip). If an activity requires movement on multiple axes, it can only be completed in steps. The prosthesis wearer must learn and practice in order to use the individual axis modes and the directions of movement correctly.

Arm nerves in the chest
The next generation of prostheses is to allow movements to be combined and to be executed intuitively and fluidly. One possible method for this is targeted muscle reinnervation (TMR). For this, nerve ends that formerly led to the arm are transplanted to a muscle that lost its function after the amputation. When an entire arm was lost, for example, the arm nerves end in the individual segments of the chest muscle.

When the patient tries to move the arm that is no longer there – the phantom arm – his or her chest muscles are stimulated. A myoelectric prosthesis with sensors above the chest muscle segments recognizes these signals and translates them into motion. TMR has been used with patients since 2012, and is especially used for high arm amputations. It requires a complicated operation.


A prosthesis that can learn
Pattern recognition eliminates the need for an operation. The prosthesis is worn like a normal myoelectric prosthesis. The difference is that instead of two sensors on the top and bottom of the arm stump, a cuff of eight sensors encircles the remaining arm. The integrated electronics also have greater processing capacity. After the prosthesis is put on, the user goes through a calibration program.

The prosthesis is connected to a computer, and the user is guided through the program steps. The user completes a wide range of movements with the phantom hand that are proscribed by the software, such as the pincer grip. The muscle activity in the arm is read during this calibration. While the various phantom movements are completed, the program learns to differentiate between the different signal patterns caused by the movements. At the end, this data is uploaded to the prosthesis, and the wearer can begin using it.

Placement challenge
In the future, calibration training with the computer should be possible with the prosthesis alone. Research is still being conducted into how the muscle signals can be recognized and differentiated most effectively. One challenge is the placement of the prosthesis stem with the sensor cuff. The latter must be placed in exactly the same place on the skin each time the prosthesis is put on to ensure good signal detection.

One way to solve this problem is implanted electrodes, explained Christian Hofer, head of the pattern detection project at Ottobock. However, his company is still looking for a non-invasive solution.

Power management
Another problem is the processing capacity in the prosthesis in connection with power management. Smartphone processors would be powerful enough, but use too much electricity.

The batteries in the prosthesis are primarily needed to power the motors that execute the arm and hand movements. It goes without saying that they should be powerful enough. Current myoelectric prostheses execute a pincer grip with a pressure of roughly 7 kilograms per square centimeter. A key can be held with a pressure of 5 kilograms per square centimeter. A prosthesis should not drop any objects.

But the battery in a prosthesis should last for at least one day. Ninety-five percent of the power in the batteries is used for the motors, and only 5 percent is used for the processor. The processor currently only has about one tenth the capacity of a smartphone processor, but uses considerably less electricity. The power-saving capabilities of mobile phone chips are constantly improving, however. So it may be possible to use them in prostheses in the future. “We have progress on our side,” Hofer said.

High expectations
The researchers at the Viennese technology center are trying to meet the high expectations for a simpler, more intuitive prosthesis control system. But there are some individual differences in the use of the pattern-detection prosthesis. In general, “People with greater body awareness have an easier time operating a prosthesis. For example, athletes,” Hofer said.

It may still be a few years before a myoelectric prosthesis with pattern recognition is introduced onto the market. Further research is necessary, and the prosthesis must then be adapted for production, and then comes the lengthy approval process. Production cannot start until this is completed. When asked when he thinks the first prostheses with pattern recognition will hit the market, Hofer was cautious: “The year 2016 at the earliest.”