By: Kameron Sprigg
In the USA alone, there are an estimated 1.6 million people living with limb loss. While current standard of care prosthetics currently increase mobility and quality of life, new research is looking to develop improved methods. Specifically, Neurally Integrated Prosthetics (NIPs) have become an area of interest[2-12]. These are connected to the body’s nervous system, and use biological signals to elicit controlled movement[2-5]. Effective NIP models have been developed, however 23% of subjects have rejected them due to discomfort, inaccurate performance, or pain .
In practice, there are two approaches to NIP development, and each is tailored to different situations. These include Brain Machine Interfaces (BMIs)[2,4-6], and Peripheral Interfaces (PIs)[7,8,10], which are systems that connect to local nerve endings. While the most obvious population that would benefit from these are those who have lost a limb, this technology can be applied to anybody who has lost limb function. This includes those who have suffered from stroke, spinal cord injury, or other nerve damage.
Both methods use similar principles. In order to build a responsive prosthetic, there must be sensors, such as electrodes or electroencephalograms (EEGs). Next, receivers and algorithms need to work together to interpret the incoming data from sensors. Finally, the prosthetic must be able to perform complex movements determined by the initial biological signals. Principally, where these vary is in the location of the sensors and specific use of algorithms. BMIs typically use EEGs that are located in the brain, acting similar to a ‘net’ over the motor region. This collects neural impulses, and wirelessly transmits the data to the receiver. Peripherally connected systems generally use multiple electrodes that are implanted directly at the site of amputation[7,10]. These then decode multiple signals from different neuro-muscular fibres to determine the intended movement.
PIs are used in amputee patients, and detect the signals of terminal nerves. This takes advantage of the decoding naturally performed by the body, while also eliminating ‘noise’, which refers to impulses from the motor cortex intended for other regions of the body. One of the most recent developments for this method comes from a study by Pasquina et al., where 8 electrodes were implanted directly into the muscles near the prosthetics attachment point. These electrodes then wirelessly transmitted data from both superficial and deep muscles to the prosthetic. Because the electrodes are implanted, the data that they transmit is more accurate and the prosthetic can be worn with greater comfort, since the prosthetics position can be shifted as necessary without compromising function. Additionally, the subject reported a more natural feel to daily interactions, such as while cooking or handling coins.
One of the major challenges with this type of model is electrode degradation. There must be a balance between protecting the sensor, while maintaining the ability to receive incoming signals from the nerves. A novel way of solving this is through the development of regeneration inducing scaffolds. These stimulate nerves to grow around the electrode, keeping surrounding tissue healthy and capable of conducting signals efficiently. This also minimizes the risk of the electrode degrading, because the myelin sheathe (nerve shell) is made of fats. This means it can act as a shield for both the electrode and nerve when the nerves grow as a scaffold surrounding the electrode. Preventing both electrode degeneration, and pain/inflammation associated with heavy metals interacting with the bodies sensitive physiology has potential to make PIs highly effective.
The other method of NIP development – BMIs, directly interprets motor signals from the brain. Typically, an EEG is used to measure neural impulses[2,4], and then computer software interprets the data based on an algorithm[8,10]. The algorithm calculates the probability of specific actions being intended based on the sequence of the nerve impulses. Another means of measuring brain impulses is the Utah Electrode Array, where very small electrodes are inserted directly into the brain. The advantage of this more invasive technique is that it allows for up to 17 different degrees of freedom (DoF) to be measured at once. One DoF could be pronating the wrist, while another could consist of closing the hand. However, this is the most susceptible to degradation of all methods, as it becomes corroded by surrounding biological material, leading to a steady decline in function.
One of the potential benefits to BMIs is that the EEG can be modified to not only receive data from the brain, but also to transmit data back into the sensory region based on receptors in the prosthetic[3,6,11,12]. This effectively allows the subject to feel their interactions, making the entire system more natural and responsive.
The final perk of using BMIs, is that by using only the brain to receive signals, the technology can be applied to those who have lost control of their peripheral nervous system[2,4,6] from events such as stroke, or spinal cord severance. When used this way, a machine separate from the patients is controlled remotely from an EEG, allowing the patient to manipulate their environment once again.
Both methods require that the subject learn to use the prosthetic, typically involving months of training before smooth and accurate use becomes second nature for the subject[2,3,6-8]. This is because the prosthetic is not a part of the body, and uses different mechanisms compared to its biological counterpart. The challenge here is to reduce the adaptation period. By examining both the brains plasticity, and how artificial intelligence can interact in conjunction with the brain. Carmena determined that maintaining the circuits’ stability is the most important factor to accelerate training. Additionally, for artificial intelligence to assist in the process, it must determine its own success rate in predicting the brains intent, while also updating its parameters frequently to match these findings[3,8].
The next steps to developing this technology are both in biological-artificial connections, and in the prosthetics themselves. PIs can be used in patients who have undergone amputation, while BMIs can be used to assist patients who have otherwise lost function. Together, the development of these both can provide greater quality of life and mobility for patients of multiple disabilities.
 Ziegler-Graham K, Mackenzie EJ, Ephraim PL, Travison TG, Brookmeyer R. Estimating the prevalence of limb loss in the United States: 2005 to 2050. Arch Phys Med Rehabil 2008;89(3):422–9.
 Hochberg, L. R., Bacher, D., Jarosiewicz, B., Masse, N. Y., Simeral, J. D., Vogel, J., . . . Donoghue, J. P. (2012). Reach and grasp by people with tetraplegia using a neurally controlled robotic arm. Nature, 485(7398), 372-375.
 Carmena, J. M. (2013). Advances in Neuroprosthetic Learning and Control. PLoS Biology PLoS Biol, 11(5).
 Tankus, A., Fried, I., & Shoham, S. (2014). Cognitive-motor brain–machine interfaces. Journal of Physiology-Paris, 108(1), 38-44.
 Rieth, L., Franklin, R., Tathireddy, P., Sharma, R., Williams, L., Tenore, F., . . . Solzbacher, F. (2011). High channel-count neural interfaces for multiple degree- of-freedom neuroprosthetics. Procedia Engineering, 25, 1365-1368.
 Yanagisawa, T., Hirata, M., Saitoh, Y., Kishima, H., Goto, T., Fukuma, R., . . . Yoshimine, T. (2010). Prosthetic arm control by paralyzed patients using electrocorticograms. Neuroscience Research, 68.
 Pasquina, P. F., Evangelista, M., Carvalho, A., Lockhart, J., Griffin, S., Nanos, G., . . . Hankin, D. (2015). First-in-man demonstration of a fully implanted myoelectric sensors system to control an advanced electromechanical prosthetic hand. Journal of Neuroscience Methods, 244, 85-93.
 Fariman, H. J., Ahmad, S. A., Marhaban, M. H., Ghasab, M. A., & Chappell, P. H. (2015). Simple and Computationally Efficient Movement Classification Approach for EMG-controlled Prosthetic Hand: ANFIS vs. Artificial Neural Network. Intelligent Automation & Soft Computing, 21(4), 559-573.
 McLean L. The early history of myoelectric control of prosthetic limbs (1945–1970). In: Scott RN, editor. Powered upper limb prostheses. Berlin/Heidelberg:Springer; 2004. p.1–15.
 Srinivasan, A., Tahilramani, M., Bentley, J. T., Gore, R. K., Millard, D. C., Mukhatyar, V. J., . . . Bellamkonda, R. V. (2015). Microchannel-based regenerative scaffold for chronic peripheral nerve interfacing in amputees. Biomaterials, 41, 151-165.
 Patient With Paralysis ‘Feels’ Physical Sensations With DARPA Neural Technology. (2015, October). Health Care Updates, 34.
 Venkatraman, S., & Carmena, J. M. (2011). Active Sensing of Target Location Encoded by Cortical Microstimulation. IEEE Trans. Neural Syst. Rehabil. Eng. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 19(3), 317-324.
 Silverthorn, D. U., & Johnson, B. R. (2015). Human physiology: An integrated approach. San Francisco: Pearson/Benjamin Cummings.