OTOtherJournal of neuroengineering and rehabilitation2017

Resolving the effect of wrist position on myoelectric pattern recognition control.

Adenike A Adewuyi, Levi J Hargrove, Todd A Kuiken

PMID 28472991

WHAT IT FOUND

For EMG control of hand grasps, adding wrist position data improved accuracy, especially with a model that handles non-linear patterns.

Simulating other wrist positions from neutral-position data also improved accuracy, but only in offline tests.

Key findings

01The non-linear neural network classifier showed a 43% relative reduction in error for extrinsic muscle data, a 48% relative reduction for intrinsic muscle data, and a 30% relative reduction for combined data, with absolute error improvements up to 5.4%.

02Combining extrinsic and intrinsic muscle EMG data performed significantly better than using intrinsic or extrinsic muscle data alone.

03Training an LDA classifier with real data from only a neutral wrist position produced errors of 27%, 22%, and 15% for extrinsic, intrinsic, and combined muscle data; training with simulated data from all wrist positions reduced error by 48%, 54%, and 74%.

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.

Already have one?

What it does not show

The study included only six non-amputees and two partial-hand amputees, so it cannot show how other partial-hand amputees would respond. Statistical analysis was performed only on non-amputee data, so the amputee results are descriptive trends. All results were offline, with four hand motion classes and no physical prosthesis or real-time task. Training and testing data came from the same day and experimental session, so stability across days is unknown. The simulated-data method predicted each EMG feature separately, which may lose information across features. Adding too many wrist positions sometimes increased error, because labels from one position could contradict labels from another.

Declared interests

Funders listed are the National Institute on Disability, Independent Living, and Rehabilitation Research, the National Institutes of Health, the Eunice Kennedy Shriver National Institute of Child Health and Human Development, and the United Negro College Fund - Merck Graduate Dissertation Fellowship.

The easy way to misread this

Do not conclude that these methods will improve real prosthetic control for partial-hand amputees. The study was offline, used four hand postures, and performed statistical testing only on six non-amputees; the two amputees showed trends but were not tested statistically.

Read it on PubMed →