Artificial Intelligence (AI) And Biomaterials: A Perfect BandAId™
What is Artificial Intelligence?According to the Webster's dictionary definition, Artificial Intelligence (AI) is: a branch of computer science dealing with the simulation of intelligent behaviour in computers, and/or the capability of a machine to imitate intelligent human behaviour. (1)
Much has already been said about AI and, unfortunately, most of it is negative. (2) From students using ChatGPT to write term papers and college admission essays to AI self-driving cars that can crash and cause injuries (3), AI has certainly received a lot of negative press lately. As usual, however, some of this negative press is by those who are not scientists and do not understand the positive impacts AI can have, especially in medicine.
During COVID-19, we have all become accustomed to false reporting about groundbreaking science and scientists and need to be cautious of such information. We should, instead, learn for ourselves. Professor Thomas J Webster is one of those learning for himself the value of AI in improving medicine, if used appropriately.
He has been exploring how AI can improve medicine, and, in particular, improve the design, fabrication, and use of biomaterials across all of healthcare. To ensure such benefits reach the patient, Professor Webster has been commercialising AI in medicine by starting new companies in a new field he has coined 'BandAId™' (Figure 1).
The conventional way biomaterials have been developedIt is clear that very little innovation in biomaterials has occurred over the past century. In fact, as just one example, we are still implanting today the same hip implant (in terms of materials and geometry) originally designed by John Charnley in 1962. (4) Yet, we are asking more from these implants: inserting them into a wider range of patients, and expecting them to perform longer and more robustly. Furthermore, they are being challenged by an ever-growing population of antibiotic-resistant bacteria. This has got to stop.
Conventionally, whether speaking of orthopaedic devices or those used for cardiovascular applications, research for new implants has included trial-and-error in vitro and in vivo testing for better material chemistries. These are chemistries that can reduce infection, reduce platelet activation, promote appropriate tissue growth, and limit chronic inflammation.
Sometimes, research has centred on selecting the best mechanical properties for that specific chemistry for a particular medical device application. While countless start-up companies have formed from such approaches, they have not in general led to the well-oiled machine improving life expectancy. (5) Note that life expectancy was decreasing in parts of the world even before COVID-19, highlighting the failures of this conventional approach to new medical device design. (6)
Figure 1: The new field of incorporating AI into medicine Professor Webster has coined "BandAId™" The BandAId™?What has been forgotten in traditional biomaterial design is that patients will respond differently to the same implant. Why is it that the same implant inserted the same way and using the same procedures will sometimes lead to implant failure and will sometimes lead to implant success? What is different? The difference is the patient and their immune system. We know that a person's immune system gets weaker as they age.
We know that females and males have different immune responses. We know that active patients rebound from injury quicker than less active patients due to altered immune systems. And we know that two perfectly healthy individuals have different immune system responses to the same material.
So, when inserting a foreign body into a patient to rebuild bone, open up clogged arteries, straighten a spine, and more, why don't we consider the various immune systems that change from patient to patient? This is where Professor Webster's vision for AI comes in.
Professor Webster and his team have developed numerous biomaterials that use AI to understand patient history and, based on that information, predict how that patient will respond to an implant. No more trial-and-error biomaterial design. And, most importantly, unlike traditional academic research, he is translating such research into real commercial products destined to eliminate medical device failure indefinitely. Let's give some examples.
Figure 2: Quarksen's breathalyser uses AI to determine if disease treatment is working by assessing patient health and predicting how well they will respond to various treatments BandAId™ ExamplesA company Professor Webster is involved in has developed a breathalyser that can detect analytes in one's breath to diagnose everything from Alzheimer's disease, infections, sexually transmitted diseases, COVID-19, and more (Figure 2). Quarksen is revolutionising sensor use and has thus made it possible to easily track disease treatment through individuals' breath. They are incorporating AI algorithms to interpret data collected from such breathalysers and predict better treatments.
As another example, Professor Webster and his team have developed the concept of 'Nano-optimizedTM' where, based on prior implant success or failure and accompanied immune response, one can implement nanotextured features through AI to either 'turn-on' or 'turn-off' the immune system (Figure 3).
Or one can develop nanotextured features to inhibit bacteria colonisation if the patient has a weakened immune system or a history of implant infection. An example in Figure 3 is provided where tendon fixation devices were Nano-optimized to promote tendon growth, whereas today's tendon fixation devices simply mechanically fix such tendons into bone. This is perfect technology for a patient with a history of orthopaedic soft tissue injuries.
Figure 4 highlights Kalnar Technologies in which Professor Webster has used his expertise to grow sensors from implants and, by gathering information of what types of cells are attaching to implants in real-time, use AI to predict whether implant infection, chronic inflammation, or appropriate tissue growth will occur. Demonstrated in this figure is how such sensors can be created by atomic layer deposition (ALD) to possess a nanoscale topography to promote bone growth in the spine of sheep after 12 weeks.
Figure 3: Prof. Webster and his team have coined the term "Nano-OptimizedTM" which uses AI to predict how well a patient's immune system will accept or reject a medical device suggesting an optimal nanotopography to achieve its goal. In this specific example, nanotextures were placed on materials used for tendon fixation regrowing tendons off of their surface to improve implant function Figure 4: Kalnar Technologies is manufacturing sensors off of implants that use AI to predict a patient response to an implant, control patient response based on that information, and communicate such information to a clinician to eliminate implant failure for the lifetime of the patient. The example shown here includes using atomic layer deposition (ALD) to grow sensors off of implants which have a nanotexture that increases bone growth in a spinal sheep model after 12 weeksLastly, Professor Webster is also involved in a company called Truss Health which is using AI and digital health to gather thermal scans from patients after implant surgery to predict implant success or failure. Shown in Figure 5 are thermal scans from a patient who recently received a knee implant.
Thermal scans were used to assess whether infection and/or chronic inflammation are setting in since both implant failure modes will increase local temperature. Such approaches can also be helpful to assess when staples should be removed as the temperature profile should return to normal, indicating healthy tissue.
Figure 5: Truss Medical is exploring how thermal scans can diagnosis proper wound healing, infection, and/or inflammation surrounding an implant. This particular example is using thermal scans to determine if infection or inflammation (which would raise temperatures) is occurring surrounding a patient's knee implant. Further, AI is being incorporated into such thermal data to predict the chances of infection and/or inflammation surrounding medical device surgery SummaryIn summary, the four brief examples provided above demonstrate the promise of AI in medicine. Whether AI is being used to assess and predict disease treatment from a breathalyser, develop nanotextured surfaces personalised for that patient to ensure implant success, interpret implantable sensor function to predict implant outcomes, or assess thermal scans to assist in implant diagnosis, AI has a bright future in medicine. It just might be the BandAId™ healthcare needs to finally move beyond using a hip implant originally designed in 1962.
ReferencesPlease Note: This is a Commercial Profile
Artificial Intelligence MSc/PG Dip/PG Cert
FacilitiesWe have our own AI laboratory situated in Gateway House featuring cutting-edge workstations and technologies such as the Emotiv Flex Gel Sensor Kit and Emotiv PRO, Lynxmotion Hexapod robot, Turtlebots, HTC Vive development kits, 3D Scanner, 3D Printer and Lego EV3 Kits.
Mobile Robotics is taught as an option at undergraduate level as well as on the Artificial Intelligence with Robotics BSc programme. On the Artificial Intelligence MSc programme students will be exposed to the more advanced techniques.
The Institute for Artificial Intelligence (IAI) conducts research into use of computational intelligence techniques on mobile robots and encourages PhD applications in this field.
Library and learning zonesOn campus, the main Kimberlin Library offers a space where you can work, study and access a vast range of print materials, with computer stations, laptops, plasma screens and assistive technology also available.
As well as providing a physical space in which to work, we offer online tools to support your studies, and our extensive online collection of resources accessible from our Library website, e-books, specialised databases and electronic journals and films which can be remotely accessed from anywhere you choose.
We will support you to confidently use a huge range of learning technologies, including Blackboard, Collaborate Ultra, DMU Replay, MS Teams, Turnitin and more. Alongside this, you can access LinkedIn Learning and learn how to use Microsoft 365, and study support software such as mind mapping and note-taking through our new Digital Student Skills Hub.
The library staff offer additional support to students, including help with academic writing, research strategies, literature searching, reference management and assistive technology. There is also a 'Just Ask' service for help and advice, live LibChat, online workshops, tutorials and drop-ins available from our Learning Services, and weekly library live chat sessions that give you the chance to ask the library teams for help.
More flexible ways to learnWe offer an equitable and inclusive approach to learning and teaching for all our students. Known as the Universal Design for Learning (UDL), our teaching approach has been recognised as sector leading. UDL means we offer a wide variety of support, facilities and technology to all students, including those with disabilities and specific learning differences.
Just one of the ways we do this is by using 'DMU Replay' – a technology providing all students with anytime access to audio and/or visual material of lectures. This means students can revise taught material in a way that suits them best, whether it's replaying a recording of a class or adapting written material shared in class using specialist software.
Why Business Leaders Should Understand AI Alignment
Daniel Langkilde is CEO and co-founder of Kognic, the leading data platform for performance-critical applications like autonomous driving.
gettyMany experts agree that artificial intelligence (AI) is one of the most important technologies of our time, predicting that its impact could be comparable to that of previous industrial and scientific revolutions. If used properly, for example, AI could improve our lives in many ways—from safe, automated mobility to novel drug discovery.
With the continuing exponential increase in computational power and data, AI systems should become far more capable, possibly equaling or exceeding human-level performance at many intellectual tasks. As business leaders, however, it is imperative that we act now to ensure AI is aligned with human values and intent.
One way to do that is with AI alignment. As explained in a TechTarget article, "AI alignment is a field of AI safety research that aims to ensure artificial intelligence systems achieve desired outcomes. AI alignment research keeps AI systems working for humans, no matter how powerful the technology becomes."
The goal, according to Towards Data Science, is to create "a set of rules or principles that an AI can refer to, to make a decision itself, knowing that by following those rules, its action will be aligned with what humans would want." Human feedback is central to this process, and there is still much work to be done.
At its core, alignment is about agreeing on expectations. Anyone who has managed a team knows how much communication is required to align a group of people.
With the emergence of ever more powerful AI, this alignment exercise will be extended to include algorithms. Imagine playing darts but not agreeing on what the dart board looks like or how to score points. If it's not clear what the objective is, we will get fuzzy and useless feedback. The same is true for machine learning (ML) algorithms.
In other words, if the designer of an AI system cannot express consistent and clear expectations through feedback, the system won't know what to learn.
Teaching Machines to Understand What Humans WantBusiness leaders often perceive AI and ML to be domains of technology experts, but that is no longer the case. The world of business leadership is increasingly data-driven. The shifts we have seen in every department across the enterprise—from marketing, finance and operations—has felt the impact of access to more complex data. AI is no different, and it needs data to grow and provide value.
In order to function properly, though, AI products need to learn the language of human preferences. Humans interpret things differently and develop preferences based on their personal perceptions. Whether we are looking at language models that drive chat interfaces, self-driving cars or even the spam filter on your email, there is a lot of ambiguity and subtlety surrounding the process of how to treat data.
The core issue is that it is difficult to know in advance what data is necessary and how to tell your "machine" what high performance really is. AI alignment could help us apply consistent logic around user preferences, regardless of the specific application.
For example, the autonomous vehicle industry has already faced many challenges in putting ML-enabled products on the road. In its capacity to power self-driving cars, AI has not lived up to consumer expectations. The problem is not one of intent. Manufacturers of automated vehicles want to ensure that their products behave as passengers expect by not making dangerous turns, obeying traffic rules, driving on safe surfaces, etc.
The problem is that there is no single way to drive. This is partially due to the complexity of decision-making associated with driving and partially due to the fact that "programming by example" (i.E., ML) is so radically different from "programming by code."
Up until now, most AI products have been trained using unsupervised learning where we let algorithms derive how to solve tasks by observing humans. But in an application as complex as driving, developers of autonomous vehicles must ask themselves two questions: 1. How, specifically, do we want this product to behave? 2. How, specifically, do we make it behave that way?
This challenge extends to other types of automated mobility. The best way to express your intent is to review examples of how the algorithm behaves and provide feedback. Human feedback can be used to steer AI products very efficiently by shaping the evolving dataset to reflect the developers' intentions and user expectations.
Iteration Is The KeyAlignment is the way forward, and the key is to approach it with an iterative mindset. We live in a fast-changing world, and expectations evolve quickly. If you assemble a large dataset, you must expect it to evolve. The challenge now is to shape your data with this evolution in mind, which in turn informs your AI products.
Contrary to common belief, AI alignment is not actually a technology problem—it's a people problem. Ultimately, the ability of the AI system to learn the right kind of rules comes down to the ability of the product developer or service provider to express what it is that they want the product to do. This is where the involvement of business leaders is absolutely essential.
Many AI-based businesses are making bold promises about their intention to deliver great services and products, but capturing human preferences is actually more difficult than it sounds. If we don't figure out a better way to do this, we will see a lot of disappointment in the next few years and it's going to be very difficult to realize the potential of AI.
It's in our collective interest to get this right. If business and technology leaders can collaborate closely on alignment, it will help to create better products and benefit humanity as a whole.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
