Part II: The Existing Practical Applications of Neuroscience & AI
In Part 1, we discussed how “the increased interest in neural networks, more specifically deep neural networks, and AI has greatly benefited various technological fields, these techniques are also expected to help further advance scientific fields, such as the neuroscience community.”?
The previous edition (Part 1) also covered that, to help simplify the ambiguous and overwhelming marriage between Neuroscience and AI,, we should first categorize the space into two overarching areas: Theoretical and Practical applications. In my case, the connection between neuroscience and AI mostly lies between Neural Oscillations (a.k.a. brainwaves) and Neural Networks (ML/DL). Please note that neuroscience is a huge field and neural oscillations are one part of the field of neuroscience. However, much of what I research and write about can be thought of and applicable towards the rest of neuroscience as well.?
Furthermore, in Part 1 we looked at the current and projected theoretical applications of neuroscience (more specifically in my niche, neural oscillations/brainwaves) and AI. In this edition, we will explore the practical applications. Unlike the research and the academic communities, which primarily seek applications of theoretical nature, such as conducting research to discover, explore, and explain anything and everything, the practical approach mainly seeks applications of practical nature, such as products or services to be marketed and sold. But the practical approach is not only about tech and/or retail products and services. It also includes practical applications in the healthcare industry. Let us now dive into some of the existing practical applications in the field of neuroscience and neuroimaging in conjunction with AI.
The practical applications of Neuroscience and AI can perhaps be further categorized into two main spaces: The existing applications and the future-looking applications. In turn, the existing applications can be categorized even further into two major areas: 1) Brain Computer Interfaces (BCI) and 2) Neuroimaging. Each of these existing areas include many sub areas that have been, and still are, developing rapidly and extensively.
BCI (Brain Computer Interface) is one of the most exciting disciplines when it comes to the practical applications of Neuroscience and AI. These practical applications, both in hardware and in software, are designed to create direct communications between the human and/or animal brains and external devices, such as computer software, external hardware, robotic limbs, and much more. While most people are quite unfamiliar with BCI discipline areas - unless big names in the industry announce BCI findings as their own unique products - this branch started way back in the 70s and has had constant and quite impressive developments ever since. To illustrate, BCI brain-implants in humans were first done in the 90s and have been researched and developed even further during the past decade.?
The collaboration between neuroscience and AI is having BCI applications become even more sophisticated. AI (ML/DL) models applied to EEG data (a.k.a. neural oscillations/brainwaves), as an example, is giving birth to all kinds of practical developments. For example, applying ML models to EEG raw data is helping us invent and develop jaw-dropping practical applications in robotics, cognitive state classifications, mental disorder diagnosis, mind-controlled devices (software and hardware), just to name a few of the real-world applications. Moreover, BCI applications, in conjunction with neural networks (ML/DL), can be used to monitor cognitive and affective states of individuals, which can have many practical applications. One of the most important factors that need to be considered when it comes to making BCIs more practical is the ability to improve the performance of the system on a variety of subjects. Deep neural networks are a promising candidate for this task.
Furthermore, various neuroscience-related practical applications - such as assessing cognitive workload, emotion recognition efforts, seizure detection, motor imagery, or sleep stage scoring - machine learning and deep learning have been successfully used in the above-mentioned EEG applications. However, there are still many challenges that need to be solved in order to improve the collaboration between Neuroscience and AI technologies. To learn more about BCI please refer to the links included in this edition.
Neuroimaging in healthcare is the second existing practical area where Neuroscience and AI are contributing greatly. To illustrate, streamlined diagnostic is perhaps one of the most practical applications of neuroimaging and AI that has been expanding quite a lot in recent years. One of these applications is the ability to automate the tasks that are currently performed by human experts. For example, the ability to diagnose epilepsy and/or sleep disorders using neural oscillations (EEG data) is widely regarded as a powerful tool for assessing the functional status of the brain. Traditionally, these diagnostics have been quite manual where EEG technicians collect, process, and visualize EEG data, which then are shared with physicians for proper diagnostics. However, with the use of neural networks (still in exploration mode) these diagnostic practices have the potential of becoming more and more automated and streamlined. This does not necessarily mean that deep neural networks will replace the EEG technicians but it would certainly help the technicians process larger sets of data with more accuracy and under a shorter amount of time.?
For example, analyzing raw datasets of EEG , LFP, fMRI, MRI, etc, using specific deep learning techniques, such as CNN, can help increase the level of accuracy and efficiency of data analysis, which may drastically help streamline patients’ diagnostics. To some extent, the neuroimaging technicians and physicians collaborate with neural networks towards the common goal of improving patient care.?
领英推荐
As you can imagine, there are many technical skills that, in the above-mentioned practical application cases, are needed in order to set up and run neural networks in BCIs and Neuroimaging applications. Some of these technical skills are understanding deep learning techniques, such as CNN, knowing Python, understanding how to organize and set up datasets to feed into the neural networks, and much more. A neuroimaging technician, for example, may not have the computer science skills needed. Therefore, as a workaround, a new set of niche careers has emerged, called Computational Neuroscience, a.k.a. Theoretical Neuroscience.
A computational neuroscientist (also see the Neuromatch Academy) uses mathematical models to capture neurological features to help conduct a plethora of scientific research in the field of neuroscience. Since most traditional BCI or Neuroimaging technicians may not have, or even be interested in, the technical skills needed to use deep learning in their workflows and analysis, they may choose to partner with computational neuroscientists.
Talking about the collaboration areas between Neuroscience and Neural networks, the first time that I ever publicly talked about this was in 2012 at the Smart Data Conference in San Jose, California. At the time, I was in roughly 2 years exploring and had noticed that most of the signals (in terms of data points) given to neural networks were oversimplified in terms of the neurology and biology to make sense of AI models in neuroscience. Gradually, over the years, I watched the neuroscience community develop ways to discover and test the best neuroscience data types available - e.g. EEG input data that align best with neural networks are frequency domain EEG signals, time domain EEG signals, and EEG images - some of which are documented in this book:
To sum up, one of the most existing practical applications of neuroscience in AI is using deep learning techniques, such as CNNs, in BCI researchers to invent and create brain-to-computer communication software and hardware applications. Furthermore, in Neuroimaging, AI techniques help technicians and physicians streamline and improve patient diagnostics and healthcare recommendations practices.
Note that in this edition we only covered the existing practical applications of Neuroscience and AI. We have yet to look at the future-looking applications, such as AGI and ASI. We will do that in the next edition of this Newsletter.?
If you are an aspiring student, there has never been a time like today where we can be anywhere in the world, with pretty much any background, and still be able to get into the field of computational neuroscience. If you have not checked it out, visit Neuromatch Academy and their YouTube channel (I took their 2021 summer course and have no affiliation with them).
What do you think? Can you think of any additional practical applications of Neuroscience and AI? Let me know and I will add to the list. Thanks!
Disclaimer: For over a decade, I have been researching and exploring ways of Neuroscience (more specifically Neural Oscillations) in AI/ML/DL, Robotics, etc. All writings are my own in the old fashioned way. I do not use ChatGPT or the like to generate content. All views, research, and work is my own. All information provided on Neuroscience & AI Newsletter is for general information purposes only and is the expressed opinion of myself, Nilo Sarraf, and not others. This includes (but is not limited to) my memberships, organizations, institutions, and/or employers.