TL;DR
This study demonstrates that implanted electrodes, particularly those integrated with regenerative peripheral nerve interfaces (RPNIs), offer significantly more accurate and stable control of prosthetic hands and wrists compared to traditional surface EMG. This improved control leads to faster task completion and reduced compensatory body movements, paving the way for more intuitive and functional upper-limb prosthetics.
What Problem Does This Paper Address?
Millions of people worldwide experience upper limb amputations, severely impacting their ability to perform daily activities and diminishing their quality of life. Current prosthetic control methods, primarily relying on surface electromyography (sEMG), often suffer from low signal quality, susceptibility to environmental factors (like sweat and movement artifacts), and limited functionality. This leads to challenges in achieving dexterous control, requiring frequent recalibration and often resulting in users opting for simpler, less functional devices. The paper addresses the need for more robust and intuitive control signals that can enable simultaneous control of multiple degrees of freedom (DoF) in prosthetic hands and wrists, mimicking natural limb function.
Methods
The core of this research lies in comparing the performance of implanted intramuscular EMG (iEMG) electrodes with conventional surface EMG (sEMG) electrodes for prosthetic control. Two participants with transradial amputations underwent surgery to create Regenerative Peripheral Nerve Interfaces (RPNIs) by suturing free skeletal muscle grafts to peripheral nerves. Intramuscular EMG electrodes were then implanted into these RPNIs and residual muscles.
The study involved several experimental protocols:
- Signal Acquisition: EMG signals were recorded simultaneously from both implanted iEMG and surface sEMG electrodes (both gelled and dry-domed types).
- Signal Processing: Raw EMG signals were filtered and processed in real-time.
- SNR Collections: Monthly sessions were conducted to measure signal-to-noise ratios (SNR) and maximum voluntary contraction amplitudes.
- Grip Classification: Participants controlled a virtual prosthetic hand and wrist in real-time using classifiers trained on EMG data. This was performed in both "still" (sitting) and "moving" (standing with dynamic arm movement) conditions.
- Functional Task (Coffee Task): One participant performed a standardized ADL assessment (Coffee Task) using both implanted and surface electrodes, with and without active wrist rotation control. This measured task completion time and trunk compensatory movements.
The key novelty lies in the use of RPNIs as a platform for implanted electrodes, aiming to capture signals from reinnervated muscle tissue, and the direct comparison of these implanted systems against both ideal (gelled) and practical (dry-domed) surface electrodes under realistic dynamic conditions.
Datasets
The study utilized data collected from two participants (S1 and S2) during various experimental sessions. No pre-existing public datasets were used for training or evaluation. The data collected included:
- Simultaneous EMG recordings from implanted and surface electrodes during grip classification tasks.
- EMG recordings during SNR collection sessions for signal analysis.
- Motion capture data and EMG recordings during the Coffee Task.
The study did not introduce a new benchmark dataset but rather used the collected data to evaluate performance against established metrics like accuracy and success rate.
Key Insights and Ideas
- RPNIs as a Superior Signal Source: The integration of implanted electrodes with RPNIs appears to be a highly effective strategy for capturing high-quality EMG signals, even for muscles that are no longer directly innervated in the residual limb.
- Dynamic Control is Key: The significant performance drop observed with surface EMG during dynamic arm movements highlights a critical limitation of current sEMG-based prosthetics in real-world scenarios. Implanted electrodes, in contrast, showed much greater stability.
- Wrist Rotation is Crucial for Natural Movement: The inclusion of wrist rotation as a controllable DoF dramatically improved functional task performance, reducing compensatory body movements and speeding up task completion. This underscores the importance of multi-DoF control for intuitive prosthetics.
- Signal Quality Matters: The higher SNR and lower cross-correlation of signals from implanted electrodes directly translate to more reliable and accurate classification of hand and wrist movements, even under challenging conditions.
Key Results
- Accuracy: In the "still" condition, implanted electrodes achieved per-bin accuracy rates of 82.1% (S1) and 91.2% (S2), significantly outperforming gelled surface electrodes (77.1% and 81.3%) and dry-domed surface electrodes (58.2% and 67.1%). During "moving" conditions, this advantage was even more pronounced, with implanted electrodes maintaining accuracy (78.3% and 75.1%) while surface electrodes saw substantial drops (e.g., dry-domed dropped to 12.3% and 22.0%).
- Success Rate (SR): Similar trends were observed for SR. Implanted electrodes showed smaller decreases in SR during movement compared to surface electrodes. For example, S1's SR dropped by 23% with implanted electrodes but 63% with gelled and 50% with dry-domed electrodes during movement.
- Coffee Task Performance: With implanted electrodes and active wrist rotation, the Coffee Task completion time was reduced by 41.26 seconds compared to surface electrodes. Furthermore, active wrist rotation significantly reduced trunk lateral lean, bringing it closer to normative values.
- EMG Signal Characteristics: Implanted electrodes demonstrated significantly higher EMG signal amplitudes during active movements (389% higher median) and lower cross-correlation between channels compared to surface electrodes.
Limitations and Future Work
- Small Sample Size: The study involved only two participants, limiting the generalizability of the findings. Future research should include a larger and more diverse participant group.
- Surface Electrode Comparison: While gelled electrodes were used for initial comparison, they are not practical for long-term prosthetic use. The primary comparison was with dry-domed surface electrodes, which may not represent the absolute best-case scenario for surface EMG. Future work could explore high-density surface electrode arrays.
- RPNI Specificity: While RPNIs were used as a source of implanted electrode data, the study did not specifically isolate the benefits of RPNIs versus implanted electrodes in residual muscles. Further research could directly compare these two approaches.
- Long-Term Stability: While the study suggests long-term stability based on previous work, this specific study did not extensively evaluate the long-term performance and stability of the implanted RPNI system over extended periods.
- Future Work: The authors suggest further investigation into the specific benefits of RPNIs for control, exploring high-density surface electrode arrays, and conducting longer-term studies to confirm the durability and sustained performance of implanted systems. The ultimate goal is to leverage these advancements for more intuitive and functional prosthetic limbs.
Summary updated