Advances in AI-based prosthetics development: editorial

The global population is becoming older, and there are more vascular disorders and injuries that cause amputations, so those figures are certain to go up. Bioengineers are aiming to develop prosthetic limbs that function as natural extensions of the body as a means of human–machine interaction. However, the end objective here is to provide amputees with prosthetic limbs the ability to do all daily tasks normally.

The world’s cinemas and television shows have now made fantastical stories based on this subject. Prosthetic limbs are more popular than ever before, especially in science fiction and other imaginary realms. The development of prosthetics has progressed from the ‘hook for a hand’ to ‘robotic limbs’, which are able to receive neurological signals from the brain’s sensors and then move in response to the user’s commands. It is difficult to build an intuitive robotic system that includes all the sensory components and motions of the severed limb. A group from UC Berkeley worked on a jig fitting technique for above-the-knee amputations after WWII, which contributed to the development of the quadrilateral socket. The Contoured Adducted Trochanteric-Controlled Alignment Method (CAT-CAM) socket, which evolved into the Sabolich Socket, was invented in the 1980s and ushered in yet another revolution in lower extremity socket technology¹. After perfecting the open-ended plug socket made of wood, they moved on to alternatives to the quadrilateral socket. A change in the socket-to-patient-contact paradigm was responsible for the improvement. In the past, sockets were simply square in form and did not have any specific features to accommodate muscle tissue. Modern designs support the patient’s musculature and existing limb by locking the bone structure into place and transferring weight uniformly.

In the early 1990s, prosthetic knees controlled by microprocessors were first made accessible to the public². The first prosthetic knee to be controlled by a microprocessor was the Intelligent Prosthesis, which was sold commercially. Designed to make walking with a prosthesis seem and feel more natural, it was produced in 1993 by the British company Chas. A. Blatchford & Sons, Ltd. Intelligent Prosthesis Plus, an upgraded version, was introduced in 1995. The Adaptive Prosthesis was another prosthesis that Blatchford introduced in 1998. To provide the amputee a stride that was more sensitive to variations in walking pace, the Adaptive Prosthesis included hydraulic controls, pneumatic controls, and a microprocessor.

Beginning in 2006, the Revolutionizing Prosthetics program was initiated by DARPAknown as the Defense Advanced Research Projects Agency, is in charge of creating new technology for military purposes³. The combination of better body armor and advances in combat trauma treatment has increased the survival of injuries sustained in the Iraq and Afghanistan conflicts. Blast and ballistic injuries, often inflicted by improvised explosive devices, may still affect the soldier’s limbs, even while body armor shields the internal organs. Amputation may be necessary as a result of these wounds.

Amazingly, our hands and arms are part of a far more complicated system that allows us to move around and engage with the environment. Developing an artificial limb that mimics the structure and function of a normal limb is an incredibly difficult undertaking. Concerns about comfort, physical attractiveness, natural control, and sensory input are key aspects of quality of life that must be recognized and addressed. System performance should be comparable to that of a human limb. Modularity and configuration are key features for an architecture that can accommodate transhumeral and transradial amputees undergoing shoulder disarticulation. Natural movement, speed, and dexterity are required, as are human forces (e.g. a 20 kg elbow curl and a 32 kg grasp), and the system must be able to assist with activities of daily living like holding a spoon and combing hair. It must also have 22 or more degrees of freedom⁴. The system needs to seem natural; it ought to be as big and heavy as the native limb and look as realistic as skin. Furthermore, it does not matter what kind of limb structure it has; it has to move organically with realistic movements and resistance to motion. The system’s control mechanism should mimic that of the original limb it is intended to replace. It would be better if this system was neurally integrated, using the body’s existing pathways for sensory input and motor control, rather of having controls that are not intuitive. A key area of progress in prosthetics is artificial intelligence (AI), which is a large aspect of robotics⁵. AI refers to a machine’s ability to think and act intelligently. Unfortunately, not all types of amputation allow prosthetic limbs to detect neurological impulses, which are necessary for movement. Not everyone can use these prostheses since they have damaged or dead nerve endings. Although nerve grafting has been around for a long time and allows for the replacement of damaged nerve endings, it is a very dangerous operation with the potential for disastrous outcomes.

Recent advances in digital healthcare technology and the Internet of Medical Things have facilitated substantial progress in the study of smart prostheses. Technological advancements in AI and robotics have paved the way for game-changing physical rehabilitation tools like exoskeletons and mind-controlled prostheses. Adding AI to these smart prostheses allows the algorithm to decipher electrical nerve impulses sent by the patient’s muscles, allowing for finer-grained control of the prosthesis. The capacity to design wearable technology is being profoundly influenced by additive manufacturing. Among the many promising new options for limb replacement, 3D-printed smart prostheses provide a promising design and manufacturing process. Australian researcher Troy Baverstock of Griffith University developed limbU, an add-on for smart prosthetic legs, using 3D printing principles to expand their usefulness^6,7. Due to the integrated electronics and sensors in limbU, the user of the prosthetic limb may more accurately react to both internal and exterior stimuli. For instance, limbU can do more than just carry out routine tasks; it can also track the wearer’s whereabouts, do sophisticated tasks like using a phone, and communicate with the doctor regarding the patient’s rehabilitation-related medical information.

Össur, a firm located in Iceland, is teaming up with the Alfred Mann Foundation, based in the US, to create lightweight, mind-controlled prosthetic limbs that can be powered autonomous for extended periods of time⁸. After an orthopedic procedure, Trac Patch may track the patient’s motion and activities throughout the day and provide that information to the surgeon. This guarantees that the recuperation period is attended to effectively. Additionally, there is a provision to provide the patient with support and encouragement as they recover. One innovative tool for tracking patients’ progress following orthopedic surgery is BPMpathway⁹. Patients are required to wear the BPMpro sensor throughout preoperative and postoperative rehabilitation and testing. It sends data to the app in real-time and monitors movement. The program presents the data as an animated graphic that changes in real-time, letting users compare the results visually. Users may follow along with videos and instructions as they complete exercises and tests on the app, which will aid them in their recovery journey. Additionally, the clinician may access the data via a web-based dashboard, which is immediately transferred via the cloud. Information such as patients’ stated pain levels, mobility, and compliance with physiotherapy activities are part of this. Clinicians may use the software to create individualized rehabilitation plans for patients prior to their hospital release. They do this so they can monitor their patients’ continuous healing and make course corrections as necessary. There is a two-way texting feature in the app that allows patients and doctors to talk to each other. Because every sensor has its own unique system identifier, the software that came with a certain person’s sensor is the only one that can be used to retrieve settings or send data.

The inability to perceive touch was a major drawback of conventional prosthesis. An e-dermis is a kind of electronic skin that mimics the sensation of touch and pain by using sensors embedded in a combination of rubber and fabric to simulate the activity of peripheral nerves¹⁰. The result is a plethora of new tactile experiences for the amputee. A neuromorphic model was created that resembles the human nervous system’s touch and pain receptors. This model enables the e-dermis to electronically register sensations in the same way as skin receptors would. Electroencephalography (EEG) was used to monitor the subject’s brain activity, and it was found that he could feel these feelings in his virtual hand. Transcutaneous electrical nerve stimulation was used to link the e-dermis output to the volunteer in a noninvasive manner. During a pain-detection assignment, the researchers discovered that both the test subject and the prosthesis could respond naturally and reflexively to the sensation of pain when touched a pointed item, but not when touched a spherical object.

Subtle actions, such picking up a dropped item or bringing fingers together, may be accomplished by the prosthesis user with the help of a regenerative peripheral nerve interface and machine learning algorithms. This is made feasible by enclosing the severed nerve with a little bit of muscle during the procedure, which increases the strength of the impulses. The University of Texas at Dallas’s Mohsen Jafarzadeh controls smart prosthetic hands using an EMG-based control system, using AI and deep learning¹¹. More precise and quicker limb motions are now possible with the help of the convolutional neural network, which is a huge breakthrough. In order to create a more advanced prosthesis, the system was retrained and adjusted utilizing the data collected from the wearable device. Situated in India smart prostheses powered by AI called Avocado make it possible to accomplish both light and heavy lifting, gardening, sketching, and ploughing¹². The Avocado is a wrist connection that includes a tiny mechanical adapter and gadget that might improve the current prosthetic user’s experience while reducing power consumption of a powered prosthetic hand. Developed by the Harvard Innovation Lab, BrainCo is a wearable that connects the human brain to a machine. A robotic prosthetic hand called BrainRobotics can be operated entirely with the user’s thoughts¹³. An exoskeleton suit that can double a person’s strength and mobility by a factor of ten has also been created by two Japanese robotics companies, Cyberdyne and Ekso Bionics¹⁴.

Research into certain algorithms and their uses in different prosthetics is crucial for moving the field of AI towards more integrated prosthetic device development. Researchers exploring neural AI algorithms can use this as a compass to lead their efforts towards meaningful innovation. To better assist those with upper-limb amputees with navigational aids and object handling, prosthetic devices can make use of convolutional neural networks, which play a crucial role in processing and interpreting visual data^15,16. The capacity to predict limb movement sequences is vital for prostheses, since it allows devices to respond more accurately to the user’s desired motions. When applied to prosthetics, reinforcement learning (RL)—especially in conjunction with deep learning techniques, or Deep RL—becomes a game-changer¹⁷. Devices are able to learn from their interactions with the user and their surroundings, enhancing their usefulness through a process of trial and error. The development of prosthetics that can adjust to their users’ changing demands and the obstacles they face in their environments relies heavily on this adaptive learning. Despite their lack of legacy in the prosthetics area, generative adversarial networks present an attractive opportunity to create synthetic data that reflects real-life situations¹⁸. AI systems can be better trained for prosthetics in this way, making them more adaptable to different user interactions and environments. Motor function prosthetics are one area that seeks to use AI algorithms like RL and LSTM networks to learn complicated movement patterns in order to mimic or improve natural limb movements. Another important field of application is sensory feedback prosthetics.

Conditions affecting physiological systems of brain activity may be amenable to treatment with neuromodulation technologies. Personalized neurostimulation regimens are essential for efficient treatments and highly effective devices. Various biological and microscale events occur over time, necessitating regular recalibration of neuromodulation settings at the interface between the targeted nerve tissue and neurotechnology, also known as the human–machine link or neural interface. An expert-mediated recalibration is typically required to adapt the optimal stimulation parameters; this adds cost, reduces the device’s everyday usability and effectiveness, and causes patients to waste time and feel more discomfort when they return to clinics for device tuning. Using time-dependent variables, researchers readjusted the neurostimulation settings in actual neuroprosthetic data using Gaussian process-based Bayesian optimization (GPBO) techniques¹⁹. In order to achieve this goal, they constructed a predictive model that can adjust the neuromodulation settings in two critical cases requiring recalibration. First, they developed a model that could identify the best active locations in a multichannel electrode to perform a certain neuroprosthesis function; in this example, that function was to account for the variability in the location of evoked sensations. They provide a method in the second one that can adjust the injection charge needed to achieve a functional activation of the neural circuits (such as variability in perceptual thresholds). The program can automatically adjust the settings of neurostimulation in response to variations in the location of evoked sensations and changes in perceptual thresholds over time. Because neurostimulation settings will inevitably vary over time, these results provide an automated, fast method to deal with this problem.

For prostheses powered by AI, this appears to be the uncharted territory. A case in point is the Esper Hand, it makes use of conventional electromyographic sensors to track electrical impulses in the muscles and convert them into hand gestures, grips, and other movements²⁰. On the other hand, the Esper’s AI components ‘learn’ to translate such signals more accurately over time. It gets easier to use the hand as its predictions get faster and more precise.

The Utah Bionic Leg is a prosthetic lower limb that incorporates intent detecting technology²¹. The AI of this system surpasses that of a microprocessor knee, which tracks the prosthesis’s spatial coordinates, processes the data, and makes constant modifications to keep the user walking naturally, maximize energy efficiency, and prevent falls caused by unstable situations. An enhanced AI system in the Utah Leg reads signals from the residual limb’s muscles and makes sense of them in relation to the user’s commands. More natural and intuitive motion is supported by that additional layer of data compared to a regular MPK. Smart prostheses must be able to seamlessly integrate with daily tasks. While the technology is intriguing, more and more of these wearable gadgets should be offered for desktop printing so that the average person can afford them. Smart prostheses are a boon because they aid patients in the recovery period after surgery by decreasing the amount of time and money spent in the hospital and increasing the motivation and speed of recovery of the patients. These devices allow patients and physicians to track each other’s progress throughout recuperation and respond quickly to any issues that crop up by integrating wearable sensors with smartphone apps. Connected controllers and sensors make powered prostheses an excellent substitute for a biological limb that has been amputated. They mimic the normal stride and movements of people who are not amputees and are therefore being fine-tuned and enhanced all the time. The promise of a better world for people with disabilities is borne out by the development of more advanced wearable technology that can sense things like touch and pain. Innovations in prosthetics, machine learning, and robotics hold great potential for amputees. Artificial limbs will include these technologies as their component prices decrease over time. All these new innovations are little tweaks that will make future prosthetic limbs better, eventually making them more like real AI. Although the widespread use of AI limbs is still a ways off, we may look forward to that day.

A complex set of issues spanning the technological, physiological, and practical realms arises when AI is integrated into prosthetic devices. The difficulty of creating AI that can precisely imitate human movement is a major barrier. This task necessitates complex algorithms that can understand the user’s intent in real-time and convert it into regulated, precise movements. In addition, cutting-edge methods for incorporating brain-interpretable sensory feedback—including touch, pressure, and temperature—are required to create prosthetics that provide a natural feeling to the user. It is essential that the AI can learn and adapt to the user’s changing tastes and demands without requiring frequent human adjustments, therefore its flexibility and learning capabilities are also vital. Another big problem is energy efficiency; balancing the computing demands of AI with the limited power available in a portable, wearable device is no easy feat. Finally, it is critical to make sure these AI systems are reliable and strong. They need to work perfectly in all kinds of situations and environments, so the user can trust them. If AIprostheses are to improve amputees’ quality of life, several obstacles must be overcome.

In order to properly and successfully utilize AI in prosthetics, it is vital to establish stringent regulatory monitoring, promote innovation, keep ethical standards high, and ensure accessibility. First and foremost, investing heavily in innovation and research is vital. This requires not just financial support but also the promotion of cross-disciplinary partnerships in order to increase the potential of AI prosthetics in areas like biomedical engineering, robotics, and computer science. Developing worldwide standards to ensure these devices satisfy the highest quality requirements in terms of safety, performance, and dependability is equally critical. To ensure that people of various socioeconomic backgrounds can access AI prosthetics, accessibility must also be a fundamental component of policy frameworks. Concerns like user privacy, data security, and permission should direct the creation and use of AI in prosthetics. Lastly, in order to guarantee that AI-enabled prosthetics are safe and effective for end-users, thorough regulatory supervision is required during their creation, testing, and implementation. To ensure that the advantages of AI prosthetics are shared fairly and ethically, it is crucial to address these concerns through well-considered policy proposals.

Due to the possibility of obsolescence and the fast pace of technical innovation, it is essential to have a comprehensive conversation about the maintenance and sustainability of AI-based prostheses. Our discussion has to cover the whole device lifetime, from design to disposal, considering not just the technical details but also the long-term effects on consumers. Managing the continual updates, repairs, and replacements of these prosthetics is vital to ensuring they continue to meet the demands of their users efficiently. By include different viewpoints, this conversation can be greatly enhanced. Integrating AI prosthetics into patient care presents both potential and obstacles, which can be better understood with the help of healthcare providers’ insights. When thinking about the ethical consequences of this technology, especially in relation to concerns of justice, accessibility, and the effects on users over the long-term, ethicists might offer insightful commentary. First and foremost, by adding patients’ voices, we gain crucial first-hand experiences that demonstrate the real-world consequences of AI-based prostheses on people’s life. The development and implementation of AI-based prostheses can be better informed by including these varied viewpoints, which will provide a thorough grasp of the prosthesis’ sustainability, ethical considerations, and lived experiences.

Despite the industry’s rapid development, it still has obstacles in terms of accessibility and affordability that must be resolved before it can realize its maximum potential. Prosthetics have become more affordable due to advancement in 3D printing, but high-tech gadgets are still out of reach for most people. The high cost of prosthetic devices is a result of the extensive research, clinical studies, technological components, and skill that go into their production. Processors play a key role in prosthetic devices; nevertheless, chip costs have skyrocketed due to a scarcity that has persisted since the epidemic struck. It has been, without a question, an incredible journey from prosthesis controlled by the body to those controlled by the muscles and now by the mind. Even though AI prosthetics are still in their infancy, there is optimism that one day they may be able to mimic a natural limb’s whole range of motion. Many AI-based initiatives are still in the prototype stage and have not yet reached the commercialization stage. The next great prosthetic limb, both practical and inexpensive, will be developed by researchers with the support of national governments, industrial units, and investors.

Ethical approval

Not applicable.

Consent

Not applicable.

Sources of funding

Not applicable.

Author contribution

S.C.: writing original draft and editing; T.B.E.: writing and editing and supervision.

Conflicts of interest disclosure

Authors declare no conflict of interest.

Research registration unique identifying number (UIN)

Not applicable.

Guarantor

Talha Bin Emran.

Data availability statement

No new data sets generated during the study.

Provenance and peer review

Not commissioned, externally peer-reviewed.