Anush Mutyala is an ambitious high school student with a passion for exploring the interdisciplinary field of brain-computer interfaces (BCIs). His interest was sparked by work presented by companies like Neuralink and has evolved into a quest to make an impact on the realm of neural implants.
Anush's recent project, titled "Enerspike: A Novel Approach to Motor Imagery Decoding on Neuromorphic Hardware”, will be presented at the Canada-Wide Science Fair. This project examines decoding neural signals into motor intent, with the goal of improving brain-computer interfaces and enabling paralyzed patients to better interact with the world around them. Central to his project is the use of Applied Brain Research's Legendre Memory Units (LMUs), a neural network that excels at encoding information over time.
We reached out to Anush to learn about his project and his future aspirations for the innovative field of BCI technology.
My name is Anush. I'm a seventeen-year-old grade eleven student residing in Brampton. My exploration of neuroscience, machine learning, and hardware engineering began in ninth grade when I was first introduced to neurotechnology. It was my introduction to companies like Neuralink, especially their white paper on a 3000 electrode system which was able to record thousands of individual neurons, that drew me in. I was inspired to join this fast-paced evolution, hoping to make my own contributions.
Currently, I'm focused on the domain of invasive neural implants. My goal is to use low-power electronics to develop superior neural implants that eliminate the need for frequent battery replacement surgeries and various dependencies like cables and wiring. I'm working to help advance completely wireless and power-efficient neural implants.
Invasive electrical neural implants is a high-growth area. Numerous companies such as Neuralink, Synchron, and Paradromics are investing substantial resources into this sector, aiming to record the activity of thousands of neurons at single cell resolution and decode specific tasks. My project specifically focuses on neural implants in the context of motor imagery, which is the task of decoding imagined movements.
The objective is to decode specific neural signals into motor intent.
The concept of thought-to-movement prediction, a common application of brain-computer interfaces, has been around for decades. However, it typically involves wired implants, which are restrictive for patients. While exploring ways to liberate BCI patients from wired implants, I stumbled upon a lecture at the University of Waterloo by Chris Eliasmith, the CEO of ABR.
This lecture introduced me to Legendre Memory Units (LMUs) and neuromorphic computing platforms such as the Loihi chip. The whole space of neural implants is a perfect domain for the application of spiking neural networks and neuromorphic hardware. It's an elegant solution because you're processing biological neural activity with artificial neurons. So, this was an exciting pathway, leading me to explore different tools like NengoDL and play around with the LMU.
That grew into the project I'm working on now that involves the LMU and other components to make this type of end-to-end spiking pipeline that can be deployed on neuromorphic chips, which can realize large power efficiency improvements.
In my research, I achieved roughly 240 times better power efficiency—2.5 orders of magnitude—compared to the existing standard.
I accomplished this by comparing a traditional non-spiking recurrent neural network with the spiking LMU, both trained to interpret motor thought processes and translate them into hand or foot movements. By leveraging ABR's technology, I demonstrated the potential applicability of this technology not only in brain-computer interfaces, but also in the space of biomedical implants in general
Chris Eliasmith's lecture provided an excellent overview of what an LMU is. It is the perfect recurrent neural network to be applied in the context of decoding information from noisy signals, like your brainwaves. Research from the Computational Neuroscience Research Group (CNRG) has demonstrated that the LMU, which provides an optimal solution for memory representation, can be implemented on neuromorphic chips. This contrasts with models like transformers or LSTMs, which aren't able to align with the architecture of neuromorphic chips.
In my project, I'm attempting to analyze a specific time frame of someone thinking about motor movements, which involves time series data. This means we're observing changes in neural activity over time.
Given the LMU's design to optimally encode memory, it was a logical choice for predicting motor movement based on a historical window of neural activity.
It's exciting to be given the opportunity to go to the national level and present this work because it validates all the work and time that has been put into the space over the years. Knowing that others recognize the potential impact of this technology is truly motivating.
As I prepare to head to Edmonton to pitch my technology, I'm most looking forward to meeting peers who share my interest in low-power electronics and neurotechnology. Given its specific nature and numerous facets of neurotechnology, I'm excited to engage with others in this field. Moreover, this will be an excellent chance to connect with industry experts and potential mentors who could help drive my research forward and increase the impact I can make.
In the short term, my aspiration is for neural implants to become widely accessible and effectively address various neurological conditions, both physical and psychological. On one side of the spectrum, we aim to restore mobility to individuals who have lost motor control. On the other end, researchers are using neural implants to potentially restore memory to Alzheimer's patients. There is a wide array of applications where this technology can make a significant difference.
The long-term goal is to use brain-computer interface technology to enhance the human operating system, essentially augmenting our intelligence. Ultimately, I want not only for those with disabilities to benefit from this technology but for it to form the foundation of human achievement in the years to come.