Determining grasp selection from arm trajectories via deep learning to enable functional hand movement in tetraplegia.
IMU
Inertial measurement unit
Machine learning
Neural networks
Neuromuscular stimulation
Spinal cord injury
Wearable
Journal
Bioelectronic medicine
ISSN: 2332-8886
Titre abrégé: Bioelectron Med
Pays: England
ID NLM: 101660849
Informations de publication
Date de publication:
2020
2020
Historique:
received:
21
03
2020
accepted:
21
07
2020
entrez:
1
9
2020
pubmed:
31
8
2020
medline:
31
8
2020
Statut:
epublish
Résumé
Cervical spinal cord injury severely affects grasping ability of its survivors. Fortunately, many individuals with tetraplegia retain residual arm movements that allow them to reach for objects. We propose a wearable technology that utilizes arm movement trajectory information and deep learning methods to determine grasp selection. Furthermore, we combined this approach with neuromuscular stimulation to determine if self-driven functional hand movement could be enabled in spinal cord injury participants. Two cervical SCI participants performed arbitrary and natural reaching movements toward target objects in three-dimensional space, which were recorded using an inertial sensor worn on their wrist. Time series classifiers were trained to recognize the trajectories using either a Dynamic Time Warping (DTW) algorithm or a Long Short-Term Memory (LSTM) recurrent neural network Participants were able to consistently perform arbitrary two-dimensional and three-dimensional arm movements which could be classified with high accuracy. Furthermore, it was found that natural reaching trajectories for two different target objects (requiring two different grasp types) were distinct and also discriminable with high accuracy. In offline comparisons, LSTM (mean accuracies 99%) performed significantly better than DTW (mean accuracies 86 and 83%) for both arbitrary and natural reaching movements, respectively. Type I and II errors occurred more frequently for DTW (up to 60 and 15%, respectively), whereas it stayed under 5% for LSTM. Also, DTW achieved online accuracy of 79%. We demonstrate the feasibility of utilizing arm trajectory information to determine grasp selection using a wearable inertial sensor along with DTW and deep learning methods. Importantly, this technology can be successfully used to control neuromuscular stimulation and restore functional independence to individuals living with paralysis. NCT, NCT03385005. Registered September 26, 2017.
Sections du résumé
BACKGROUND
BACKGROUND
Cervical spinal cord injury severely affects grasping ability of its survivors. Fortunately, many individuals with tetraplegia retain residual arm movements that allow them to reach for objects. We propose a wearable technology that utilizes arm movement trajectory information and deep learning methods to determine grasp selection. Furthermore, we combined this approach with neuromuscular stimulation to determine if self-driven functional hand movement could be enabled in spinal cord injury participants.
METHODS
METHODS
Two cervical SCI participants performed arbitrary and natural reaching movements toward target objects in three-dimensional space, which were recorded using an inertial sensor worn on their wrist. Time series classifiers were trained to recognize the trajectories using either a Dynamic Time Warping (DTW) algorithm or a Long Short-Term Memory (LSTM) recurrent neural network
RESULTS
RESULTS
Participants were able to consistently perform arbitrary two-dimensional and three-dimensional arm movements which could be classified with high accuracy. Furthermore, it was found that natural reaching trajectories for two different target objects (requiring two different grasp types) were distinct and also discriminable with high accuracy. In offline comparisons, LSTM (mean accuracies 99%) performed significantly better than DTW (mean accuracies 86 and 83%) for both arbitrary and natural reaching movements, respectively. Type I and II errors occurred more frequently for DTW (up to 60 and 15%, respectively), whereas it stayed under 5% for LSTM. Also, DTW achieved online accuracy of 79%.
CONCLUSIONS
CONCLUSIONS
We demonstrate the feasibility of utilizing arm trajectory information to determine grasp selection using a wearable inertial sensor along with DTW and deep learning methods. Importantly, this technology can be successfully used to control neuromuscular stimulation and restore functional independence to individuals living with paralysis.
TRIAL REGISTRATION
BACKGROUND
NCT, NCT03385005. Registered September 26, 2017.
Identifiants
pubmed: 32864392
doi: 10.1186/s42234-020-00053-5
pii: 53
pmc: PMC7449026
doi:
Banques de données
ClinicalTrials.gov
['NCT03385005']
Types de publication
Journal Article
Langues
eng
Pagination
17Informations de copyright
© The Author(s) 2020.
Déclaration de conflit d'intérêts
Competing interestsCEB holds patents in related fields and is the founder of Neuvotion, LLC, a company focused on movement restoration.
Références
Lancet. 2017 May 6;389(10081):1821-1830
pubmed: 28363483
Spinal Cord. 2008 Apr;46(4):255-74
pubmed: 17846639
Sensors (Basel). 2017 Jun 01;17(6):
pubmed: 28587178
Sensors (Basel). 2019 Sep 04;19(18):
pubmed: 31487894
Nature. 2016 Apr 13;533(7602):247-50
pubmed: 27074513
J Electromyogr Kinesiol. 2014 Dec;24(6):795-802
pubmed: 25287528
Spinal Cord. 1999 Aug;37(8):560-8
pubmed: 10455532
J Hand Surg Am. 2008 Apr;33(4):539-50
pubmed: 18406958
PLoS One. 2020 Jan 13;15(1):e0227039
pubmed: 31929544
J Neuroeng Rehabil. 2017 Mar 11;14(1):20
pubmed: 28284228
Data Min Knowl Discov. 2017 Jan;31(1):1-31
pubmed: 29104448
Neural Comput. 1997 Nov 15;9(8):1735-80
pubmed: 9377276
Bioelectron Med. 2019 Nov 1;5:19
pubmed: 32232108
J Neuroeng Rehabil. 2018 Jun 26;15(1):57
pubmed: 29940991
J Am Acad Orthop Surg. 2004 Mar-Apr;12(2):72-9
pubmed: 15089080
World J Orthop. 2015 Jan 18;6(1):8-16
pubmed: 25621206