Abstract Information

O-169

The EEG-controlled noninvasive MoreGrasp neuroprosthesis - decoding of multiple natural single limb movements and multipad-electrodes for closed-loop grasp pattern control

1Rupp R, 1Schneiders M, 1Hessing B, 2Murray-Smith R, 2Ramsay A, 3Luzhnica G, 3Veas E, 4Schwarz A, 4Pereira J, 4Ofner P, 4Pinegger A, 4Mueller-Putz G
1Heidelberg University Hospital - Spinal Cord Injury Center, Heidelberg, BW, Germany; 2University of Glasgow - School of Computing Science, Glasgow, Scotland, United kingdom; 3KNOW Center, Graz, , Austria; 4Graz University of Technology - Institute of Neural Engineering, Graz, , Austria

Objective:
Motor neuroprostheses based on functional electrical stimulation (FES) can restore permanently lost functions in people with high spinal cord injury (SCI), specifically of the grasping. Noninvasive FES using surface electrodes represents a less complex, easy-to-apply alterative to invasive approaches [1]. An EEG-based Brain-Computer Interface (BCI) may be used to control grasping by imaginations or attempts of movements [2]. While the feasibility of the combination of BCI with FES was already shown in single case studies [3, 4], intuitive EEG-based BCI-neuroprosthesis control is still missing. As the output of noninvasive BCIs is mainly digital, autonomous, closed-loop grasp generation is needed. Additionally, methods for supporting the end users in electrode placement need to be established.
The European MoreGrasp consortium tries to overcome these problems by realization of an individualized, sensorized forearm sleeve with integrated multi-pad electrodes and the implementation of a BCI based on the decoding of single limb movements.

Design/Method:
In two high-resolution EEG studies with 15 able-bodied subjects each, the decoding classification accuracy of 6 single joint movements of the same arm and of 3 different grasp types of the same hand were determined. In both studies, the motor-related cortical potentials (MRCPs) in a narrow 0.3 to 3 Hz band were investigated [5, 6]. Following the protocol of these 2 studies, 2 subsets of movements were classified in 5 participants with high cervical SCI (Neurological level: C3 - C5).
Two sets of multi-pad electrodes were developed: 1) a stackable screening electrode matrix consisting of 15 (5 x 3, HxW, 6.3 x 3.8 cm) electrodes (diameter 7mm, inter electrode distance of 2.5 cm) made of conductive silicone, and 2) a personalized forearm silicone sleeve with 64 electrodes and two inertial measurement units (IMUs) for wrist rotation angle measurement. A tablet computer based software for determination of the most selective and robust electrode positions was developed.

Results:
The 1st BCI-study revealed a classification accuracy of 37% (chance level 16.7%), with classifier sources mainly in premotor and primary motor areas. The 2nd study showed that grasps can be decoded from MRCP features (binary classification of 74% grasp vs. grasp). Experiments with SCI showed a classification accuracy of 53 % (subset 1) and 57 % (subset 2).
The test results of the multi-pad electrodes in 3 able-bodied subjects and 1 end user with SCI reveal that not only a semi-autonomous quantification of the degree of denervation is possible, but also robust electrode positions for palmar or lateral grasps and electrode switching strategies can be defined for generation of a wrist-rotation-angle-independent grasp force.

Conclusion:
The studies show that it is possible to detect single movements of the same arm from the EEG, either single joints or different grasp patterns. The multi-pad concept of the MoreGrasp grasp neuroprosthesis helps to overcome major challenges of noninvasive grasp neuroprostheses for every-day use. The system is currently tested in a proof-of-concept study at end users’ homes.

Support:
This project is supported by the EU ICT Programme Project H2020-643955 MoreGrasp.


Content

You Must Be Logged In


This Reminders tab will display for your scheduled lectures 15 minutes before start time and for new messages waiting.
Top