In order to truly translate hand prostheses research into real-world applications, and “close the loop” for a bidirectional prosthesis, Advancement in sensory feedback is required. One method for conveying non-invasive sensory feedback to users involves transcutaneous electrical nerve stimulation (TENS). This technique involves the delivery of electrical impulses to the skin’s surface, thereby generating action potentials in the underlying neurons and creating tactile sensations. The benefits of TENS include its non-invasiveness, which reduces the risk of infection and complications associated with surgical implants. Additionally, TENS allows for adjustable stimulation parameters, enabling customization of sensory feedback for individual users. This flexibility can accommodate different levels of sensitivity and adapt to changes over time. Moreover, TENS is relatively low-cost and has a small footprint making it a practical option for widespread implementation. Furthermore, TENS has been shown to have therapeutic benefits, such as pain relief, which can be particularly beneficial for amputees experiencing phantom limb pain. Overall, the integration of TENS into prosthetic devices represents a significant step forward in improving the functionality and user experience of artificial limbs, bringing us closer to the goal of restoring natural sensory feedback in prosthetics.
Project description:
A critical area of development is enabling users to perceive the sensation of movement, an enhancement that would significantly contribute to the diversity and richness of sensory feedback. The ability to detect movement is crucial for identifying slippage. Currently, without such sensory feedback, individuals using prosthetics tend to apply excessive force to ensure they do not drop objects—a practice that poses risks, especially when handling delicate items. If users could detect slippage, they would gain confidence in their ability to ascertain the necessary amount of force required to manipulate various objects safely and effectively.
Nonetheless, challenges persist with traditional sensors embedded in prosthetic limbs. While these sensors are adept at measuring applied pressure, they fall short in detecting sliding motions, a critical component for advanced, responsive feedback systems. However, a groundbreaking development known as E-skin (ACES), pioneered by Benjamin Tee, offers a promising solution. This technology features an array of sensors that transmit information asynchronously, showing considerable potential in identifying sliding movements.
By pairing the E-skin technology with a multi-electrode stimulation array, it becomes possible to employ current steering to convey the sensation of sliding to the user. This synergy between sensors array and multi-electrode TENS would allow for the generation of a richer sensory feedback system, enhancing the close-looped system of bidirectional prosthesis. The implications for enhanced dexterity and improved quality of life are profound, marking an exciting frontier in the realm of bidirectional prosthetic.
Relevant papers:
“A neuro-inspired artificial peripheral nervous system for scalable electronic skins” Lee et al, 2019
“Restoring tactile sensations via neural interfaces for real-time force-and-slippage closed-loop control of bionic hands” Zollo et al, 2019
Activities:
- Literature review and experimental protocol design
- Decoding movement information from sensors embedded in robotic hand
- Encode the information into stimulation parameters for movement sensory feedback
- Perform statistical tests
Requirements:
Project is 50% experimental, 50% analysis.
Proficiency in programming with Python is a MUST
Experience with data analysis, as well as machine learning, is highly valued
Contact: [email protected]
Project description:
One of the primary challenges lies in the calibration process for stimulation, typically a labor-intensive task that demands substantial expertise from researchers. Furthermore, stimulation parameters are prone to variations over time, influenced by factors such as alterations in skin conductance. Consequently, it is imperative to develop an automated calibration system capable of reducing the required setup time significantly.
Highlighting a breakthrough in this realm, a recent study by Borda et al. showcased the development of a Reinforcement Learning (RL) algorithm designed to determine optimal neurostimulation parameters to re-establish sensory feedback autonomously. Their findings underscore the potential of deep learning applications as viable alternatives to traditional, more laborious methods. In light of these advancements, this project intends to delve into various optimization methodologies to expedite and enhance the calibration process as well as to improve on the quality of the sensation. Some of the investigated optimization methods would include Bayesian optimization and reinforcement learning.
Relevant papers:
“Automated calibration of somatosensory stimulation using reinforcement learning” Borda et al, 2023
“HUMAN-IN-THE-LOOP optimization of visual prosthetic stimulation” Fauvel and Chalk, 2021
“User preference optimization for control of ankle exoskeletons using sample efficient active learning” Lee et al, 2023
Activities:
- Literature review on the relevant topic
- Develop or adapt an ML algorithm for determining optimal neurostimulation parameters.
- Validate algorithm by performing TENS on healthy subjects
- Perform statistical test
Requirements:
Project is 50% experimental, 50% analysis.
Proficiency in programming with Python is a MUST
Experience with data analysis, as well as machine learning, is highly valued
Contact: [email protected]
A significant challenge in integrating TENS (for sensory encoding) with electromyography (EMG) recordings (for motor decoding) is the interference encountered in the signal during TENS stimulation. The process of administering current to the skin’s surface can be inadvertently picked up by EMG electrodes, leading to artifacts that can confuse the decoding algorithms. Therefore, it becomes crucial to devise a strategy, utilizing either analytical methods or machine learning solutions, to effectively eliminate said artefacts.
This project encompasses integrating TENS with EMG to develop an initial prototype of the bidirectional prosthesis. Subsequently, the student will undertake a thorough analysis of the EMG signals in scenarios with and without TENS stimulation, aiming to understand the nature of potential artifacts induced by such electrical stimulation. If found to be sufficiently detrimental to EMG decoding, the primary objective following is to devise methods to eliminate these artifacts, thereby ensuring the decoding process’s accuracy remains uncompromised. The student will first consider an analytical strategy, such as signal removal through Fourier transformation. If traditional methods prove insufficient, the project will pivot to more complex, non-linear deep-learning techniques, like employing a denoising autoencoder.
Relevant papers:
“A somatotopic bidirectional hand prosthesis with transcutaneous electrical nerve stimulation based sensory feedback” D’Anna et al, 2016
Activities:
- Literature review
- Artifact characterization
- Code implementation for artifact removal
Requirements:
Project is 20% Hardware, 40% experiments, 40% analysis/optimization
Proficiency in programming with Python is a MUST
Experience with data analysis, as well as machine learning, is highly valued
Contact: [email protected]
Contact
If none of the projects suit you but you are interested in optimizing tactile sensory feedback in general, please feel free to contact us to discuss potential opportunities.
Franklin Leong ([email protected])