Online mapping of EMG signals into kinematics by autoencoding.
Ivan Vujaklija, Vahid Shalchyan, Ernest N Kamavuako and 3 others
PMID 29534764WHAT IT FOUND
In a small group of able-bodied people, an autoencoder muscle-signal control moved a cursor faster and with better path efficiency than standard sequential control.
This is a lab feasibility test, not evidence for amputee prostheses.
Key findings
01In seven able-bodied subjects, autoencoder control and standard direct control completed nearly all cursor targets, with completion rates 99.4 ± 0.06% and 100%, not significantly different.
02Autoencoder control had significantly shorter completion time (3.75 ± 0.32 s) than direct control (6.91 ± 0.79 s; p = 0.004).
03Autoencoder control had higher throughput (1.19 ± 0.08 bit/s vs 0.66 ± 0.06 bit/s; p < 0.001) and better path efficiency (73.80 ± 4.98% vs 54.10 ± 2.47%; p < 0.001).
STILL TO COME
How it was doneWhat they foundWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study included only seven able-bodied subjects, so it does not show how amputees or people using prostheses would control the system. The test was a cursor task in a lab, not a real prosthetic device or functional daily task. Only two wrist degrees of freedom were tested; three simultaneous degrees of freedom were too challenging for most subjects. The autoencoder was tested with position control while the industrial direct control was tested with velocity control, so the two systems were not matched in the same control mode. Calibration did not record kinematics, and the method assumes a linear association between extracted EMG primitives and movement. The paper did not report patient outcomes, safety, long-term use, or training effects.
Declared interests
H2020 European Research Council funded the work. The paper does not report other conflicts of interest.
The easy way to misread this
Do not conclude this is evidence that autoencoder muscle-signal control improves prosthetic function for amputees. It tested only able-bodied people moving a cursor, not patients or real prostheses.