RCTJournal of neuroengineering and rehabilitation2021

User training for machine learning controlled upper limb prostheses: a serious game approach.

Morten B Kristoffersen, Andreas W Franzke, Raoul M Bongers and 3 others

PMID 33579326

WHAT IT FOUND

Neither game-based nor conventional coaching training improved EMG pattern quality or functional prosthesis use in four adults with upper limb absence.

Direct control with their own prosthesis consistently outperformed machine learning control at these training levels.

Key findings

01Training did not produce consistent improvements in EMG pattern separability or consistency for either the game or conventional coaching group.

02Functional outcomes using the machine learning prosthesis did not improve with training, and baseline performance using participants' own direct control prostheses was consistently superior.

03All four participants who completed the study used two degrees of freedom at the post-test, but none had sufficient control to add a third degree of freedom.

STILL TO COME

How it was doneWhat they found

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

Only four of eight recruited participants completed the study, with dropouts due to poor socket fit, fatigue, and scheduling issues. The training period was likely too short to detect functional improvements or robust control of multiple degrees of freedom. Participants had years of experience with their own direct control prostheses but were naïve to the machine learning system and study prosthesis, biasing the comparison. No home training was included, which may have been necessary to achieve proficiency. Socket fit issues caused one exclusion, highlighting the sensitivity of EMG control to hardware configuration.

Declared interests

The study used commercial prosthetic components from Otto Bock. The authors declare no competing interests. Funding was from the European Union's Horizon 2020 research and innovation programme.

The easy way to misread this

Do not conclude that machine learning control is inferior to direct control for experienced users. This pilot found direct control was superior, but participants had years of experience with their own devices and only brief training on the machine learning system. Longer training and home use may change this result.

Read it on PubMed →


The study

Participants
4 completed of 8 recruited
Certainty of evidence
Moderate

Browse

    Cite

    Morten B Kristoffersen, Andreas W Franzke, Raoul M Bongers, et al. User training for machine learning controlled upper limb prostheses: a serious game approach. Journal of neuroengineering and rehabilitation. 2021.

    Read the original