Tag

Tagged: data mining

Sponsored
3 years, 1 month ago

A prescription for an AI inspired MedTech industry

HealthPad
Directory:
Medical Technology
Tags:
AI
Artificial Intelligence
commentary
data mining
Medtech

 

  • A wind of change is blowing through MedTech markets
  • MedTech markets have matured and are experiencing slower growth and increased competition, which have fuelled endeavours to increase growth rates
  • Artificial intelligence (AI) techniques applied to data from existing devices have the potential to achieve this and improve care
  • Obstacles to developing AI solutions include rigid manufacturing mindsets and a dearth of appropriate talent
  • To remain relevant MedTech leaders will need to “think beyond physical products”, develop new business models, new types of investments and new approaches to R&D
  • Will a wind of change that is blowing through MedTech markets be perceived as a temporary breeze?
 
A prescription for an AI inspired MedTech industry
 
Thinking beyond physical products and the growing significance of AI in MedTech markets


A wind of change is blowing through MedTech markets, which has prompted some key opinion leaders to think beyond physical products and begin to use artificial intelligence (AI) techniques to develop value added services that bolt-on to their existing physical offerings to improve clinical care and economic efficiencies while providing access to new revenue streams.

Bryan Hanson, Zimmer-Biomet’s CEO, recently suggested that >70% of his company’s R&D spend is now being invested in data informatics and robotics. Not far behind is Stryker, another global orthopaedic corporation, which has implemented AI strategies to improve care and differentiate its offerings. Both are thinking beyond their physical products to create a suite of services derived from AI enhanced data collected from their existing devices. Such actions provide a template that can be copied by other enterprises. How long will it take for AI solutions to represent a significant percentage of MedTechs’ revenues?
 
In this Commentary

This Commentary: (i) describes the growing significance of AI, (ii) explains the difference between data mining, AI, and machine learning, (iii) illustrates AI technologies that have become an accepted part of our everyday lives, (iv) highlights technical drivers of AI solutions, (v) describes obstacles to the development of AI systems, (vi) indicates how such obstacles may be reduced, (vii) describes Zimmer’s and Stryker’s AI driven data initiatives, (viii) suggests that the Zimmer-Stryker AI template has broad potential, (ix) suggests that AI systems can breathe life into 'dead data', (x) provides an example of a company at the intersection of medical information and AI techniques, (xi) describes the origins of the phrase, ‘wind of change’, and defines the ‘winds’ driving change in current MedTech markets, (xii) reports that ~80% of B2B sales in the economy generally are digitally driven, (xiii) provides some reasons for MedTechs’ slow adoption of AI systems, (xiv) floats the idea that the future for producers is to partner with tech savvy start-ups and (xv) describes how US AI supremacy is being challenged.
 
AI: vast and fast growing
 
It is challenging for baby boomers and older millennials, who populate MedTechs’ C suites, to fully grasp the potential of AI. This is largely because their corporate careers were underway before the digital age started, and for three decades they have personally prospered from manufacturing physical devices without the help of AI.
 
A person who understands the potential of AI is Sundar Pichai, the CEO of Alphabet, one of the world’s largest tech companies. In a recent BBC interview Pichai suggested, "AI is the most profound technology that humanity will ever develop and work on . . .  If you think about fire or electricity or the Internet, it's like that, but even more profound". This suggests that Hanson is right to redirect Zimmer’s R&D spend towards AI-driven solutions. A February 2021 report from the International Data Corporation (IDC), a market intelligence firm, suggests that the current global AI market is growing at a compound annual growth rate (CAGR) of ~17% and is projected to reach ~US$554bn by 2024.
 
Data mining, AI, machine learning and neural networks

Among MedTechs’ C suites there is some confusion about data strategies and AI solutions. Many enterprises use data mining techniques on existing large datasets to search for patterns and trends that cannot be found using simple analysis. They employ the outcomes to increase revenues, cut costs, improve customer relationships, reduce risks and more. Although data mining is commonly used when working on AI projects, in of itself, it is not AI. So, let us briefly clarify.

AI is the science and engineering of developing intelligent computer programs to enable machines to provide requested information, supply analysis, or trigger events based on findings. AI creates machines that think, learn, and solve problems better and faster than humans. This is different to traditional computing, where coders provide computers with exact inputs, outputs, and logic. By contrast, AI systems can be “schooled” to carry out specific tasks without being programmed to do so. This is referred to as machine learning, which usually requires large amounts of data to train algorithms [mathematical rules to solve recurrent problems].

A critical element of machine learning’s success is neural networks, which is an AI technique modelled on the human brain that is capable of learning and improving over time. Neural networks are comprised of interconnected algorithms that share data and are trained by triaging those data: a process referred to as ‘back propagation’. In healthcare, machine learning outputs range from the ability to recognise images faster and more accurately than health professionals to making in vivo diagnoses.
 
AI systems have become an accepted part of our everyday lives without us realising it
 
Most people are aware of significant AI breakthroughs such as self-driving cars and IBM’s Watson computer winning the US quiz show Jeopardy by beating two of the best players the show had produced. Lesser known, is in 2012, AlexNet, a neural network learning system, won a large-scale visual recognition contest, which previously was thought too complex for any machine. In 2016, Google’s AlphaGo, a machine learning algorithm, defeated Lee Sedol, who was widely considered the world’s greatest ever player of the ancient Chinese game Go. Most observers believed it would be >10 years before an AI programme would defeat a seasoned Go champion. Although Go’s rules are simple, the game is deceptively complex, significantly more so than chess. It has a staggering 10170 possible moves, which is more than the number of atoms known in the universe. Significantly, machine learning algorithms embedded in AlphaGo, mastered the game without any prior knowledge and without any human input. More recently Google launched AlphaGo Zero, an AI system, which can play random games against itself and learn from it. During the decade of these breakthroughs, AI systems became an accepted part of our everyday lives without us realising it. Examples include, Google searches, GPS navigation, facial recognition, recommendations for products and services, bank loans we receive, insurance premiums we are charged, and chatbots, which organizations use to provide us with information.
 
Technical drivers of AI systems

In addition to commercial drivers, AI techniques are driven by easy availability of data, an explosion in computing power and the increased use of clusters of graphic processing units (GPUs) to train machine-learning systems. These clusters, which are widely available as cloud services over the Internet, facilitate the training of more powerful machine-learning models. An example is Google's Tensor Processing Unit (TPU), which has the capability to carry out more than one hundred thousand trillion floating-point operations per second (100 petaflops). This has the potential to accelerate the rate at which machine-learning models can be trained. Further, the cloud has made data storage and recovery easier, which has motivated government agencies and healthcare institutions to build vast unstructured data sets that they make accessible to researchers throughout the world to stimulate innovation.
 
Obstacles to the development of AI systems
 
So far, we have emphasised the benefits of AI, but there are concerns that machine intelligence will accelerate at an incomprehensible rate, surpass human intelligence, and transform our reality. This is referred to as “singularity”, which has generated concerns from key opinion leaders. Nearly a decade ago, Stephen Hawking, a pre-eminent British scientist, warned in a BBC interview, that singularity “could spell the end of the human race”. More recently, Hawking’s view has been echoed by Elon Musk, founder, and CEO of Tesla and SpaceX, who suggests that AI is, “more dangerous than nuclear warheads and poses a fundamental risk to the existence of human civilization". Musk has called for stronger regulatory oversight of AI, and more responsible research into mitigating its downsides. In 2015, he set up OpenAI, a non-profit research organization, with a mission to promote and develop AI systems that benefit society. 

 

In the June 2018 edition of the Atlantic Review, Henry Kissinger, who served as national security adviser and secretary of state for two US Presidents, described the potential harms from AI by addressing the question: “What would be the impact on history of self-learning machines that acquired knowledge by processes particular to themselves, and applied that knowledge to ends for which there may be no category of human understanding?”. Singularity might be more imminent than once thought. In a book published in 2015, futurist Ray Kurzweil predicted that singularity would occur in ~2045, but a paper published in the June 2020 edition of the International Journal of Astrobiology suggests that it is more likely to occur within the next decade.

You might also like:

Robotic surgical spine systems, China, and machine learning
Overcoming obstacles to AI
 
In clinical settings there are growing concerns that complex algorithms can blur the reasoning behind specific machine interpretations and consequent actions of robotic surgical systems. As AI and machine learning develop so surgical robots are expected to become more autonomous and have the capability to make instantaneous diagnoses and pursue immediate therapies, which surgeons using the systems do not fully understand. The failure of humans to understand the workings of an AI system is referred to as an “interpretability challenge”, or more commonly, the “black-box” problem, which could impact future clinical regulations.
 
Combatting the possible dangers of AI systems not being understood by humans is a relatively new and growing research area, referred to as “Explainable AI” (XAI). XAI attempts to use AI techniques to develop solutions that can describe the intent, reasoning, and decision-making processes of complex AI systems in a manner that humans can understand. This could provide Stryker and Zimmer, and other manufacturers, a solution to potential future regulatory obstacles associated with advances in their robotic surgical systems
.
Zimmer’s and Stryker’s initiatives

In August 2021, the FDA granted De Novo marketing authorization [applicable for a new and novel device whose type has not previously been classified] for a “smart knee”, which Zimmer had developed in partnership with Canary Medical, a data analytics company. The device, called Persona IQ®, is the world's first and only smart knee cleared by the FDA for total knee replacement surgery. It combines Zimmer’s proven and trusted knee implant, Persona® The Personalized Knee®, with Canary’s proprietary sensor technology, which provides real-time feedback on how surgical implants and devices are working by generating self-reports on patient activity, recovery, and treatment failures, without the need for physician intervention and dependence upon patient compliance. The partnership is also expected to leverage Canary’s machine learning capabilities to identify further patterns in data from implants that could help clinicians catch problems, such as infections or loosening of the implants before they worsen. Persona IQ® will work together with Zimmer’s remote care management platform, mymobility® with Apple Watch®, as well as with other components of the  ZBEdge™ connected intelligence suite of currently available, and soon to be launched, digital and robotic technologies engineered to deliver transformative data-powered clinical insights, shared seamlessly across the patient journey, to improve patient outcomes. 

In January 2021, Stryker acquired OrthoSensor, a privately held technology company that makes intraoperative sensors for use in total joint replacements. Stryker expects these sensors to empower surgeons with AI-driven solutions and enhance its surgical robotic systems by eventually providing them with the capability to predict surgical outcomes. Additionally, OrthoSensor’s remote patient monitoring wearables, combined with a cloud-based data platform, are expected to significantly improve Stryker’s data analytics capabilities. According to a Stryker press release issued at the time of the acquisition, “OrthoSensor quantifies orthopaedics through intelligent devices and data services that allow surgeons and hospitals to deliver evidence-based treatments for all healthcare stakeholders. The company’s advancements in sensor technology, coupled with expanded data analytics and increasing computational power, will strengthen the foundation of Stryker’s digital ecosystem”.
 
The Zimmer-Stryker AI template has potential across MedTech

Despite Zimmer’s and Stryker’s AI-driven data initiatives to improve their respective competitive advantages and gain access to new revenue streams, few MedTechs collect, and store the data produced by their existing devices, and even fewer use such data to provide novel AI solutions. The Zimmer-Stryker template for achieving this is not limited to orthopaedics. For example, consider neuro critical care and traumatic brain injuries (TBI), which are a “silent epidemic”. Each year, globally ~69m individuals sustain TBIs. In the US, every 15 seconds, someone suffers a TBI. In England, ~1.4m people present at A&E departments each year following a head injury.

Despite extensive research, successful drug therapies for TBI have proven to be elusive. The gold standard management of the condition is to monitor intracranial pressure (ICP) and attempt to avoid elevated levels, which can cause further insults to an already damaged brain. Currently, there are no FDA approved means to identify advance warnings of changes in ICP. However, it might be possible to create an early warning of ICP crises by applying machine learning algorithms to standard physiological data produced by existing medical devices commonly used to monitor patients with TBI. This would not only provide time for interventions to prevent further trauma to critically ill patients but would also give producers access to new revenue streams.



You might also like:

MedTech must digitize to remain relevant
Breathing life into dead data

There are potentially limitless opportunities to improve care by breathing life into 'dead data'. This can be achieved simply by applying AI solutions to underutilized data from existing medical devices. The global MedTech industry is comprised of ~6,000 companies (mostly small to medium size). The overwhelming majority of these manufacture devices that produce, or could produce, patient data. These companies serve ~14 surgical specialisms each of which treat numerous conditions. For each condition there are millions of patients at any one time. For each patient, multiple devices used in therapies display real time data. Most producers are awash with dead data because they do not collect, store, and analyse these data to improve the quality of care. AI systems can change this.
A MedTech start-up at the intersection of medical information and AI techniques

A start-up, which understands the clinical and economic potential from the intersection of medical data and AI solutions is Komodo Health, which was founded in 2014. According to Web Sun, the company’s co-founder, and president, “We had a vision that integrating robust data with software solutions was the way forward for healthcare at a time when no one was doing this”. Komodo has created an AI platform, which it refers to as a "healthcare map", comprised of large-scale anonymous health outcome data from hundreds of sources.

In January 2020, Komodo announced a deal to import Blue Health Intelligence’s patient data onto its platform. Blue Health provides US healthcare claims data and actionable analytics to payers, employers, brokers, and healthcare services. The combined database charts >325m individual patient care journeys through tests and therapies at hospitals and clinics. In March 2021, Komodo raised US$220m to extend its platform to offer real-time assessments of patients’ healthcare journeys to detect disparities in the quality of care and outcomes, and to provide a basis for interventions aimed at improving outcomes and lowering costs.

The ability to introduce clinical insights into enterprise workflows potentially helps producers and providers close gaps in care journeys and address unmet patient needs. Not only are Komodo’s services designed to deliver timely interventions and alerts to improve care, but the company also records and reports the performance of specific medical products on patient cohorts. These data provide a basis to develop and market further innovative healthcare services, and novel therapeutics, which are expected to boost Komodo’s revenues.
 
A wind of change

We borrowed the ‘wind of change’ phrase used in our introduction from a famous speech made by British Prime Minister Harold Macmillan to the Parliament of South Africa on 3 February 1960 in Cape Town. Macmillan was referring to a system of institutionalised racial segregation, called Apartheid, which enforced racial discrimination against non-Whites, mainly predicated on skin colour and facial features. Despite the UK Prime Minister’s belief that in 1960, the days of White supremacy in South Africa were numbered, it took >30 years before Apartheid was ended and Nelson Mandela was inaugurated as the first Black President of South Africa on 10 May 1994. Mandela was an anti-apartheid activist and lawyer, who had spent 27 years as a political prisoner under the Apartheid regime.

A wind of change is now blowing through MedTech markets. In less than a decade, healthcare will be faced with significantly more patients, more data, more technology, more costs, more competition, and less money for producers and providers. Over the past five years, US providers’ profit margins have fallen, in Europe the gap between public health expenditure and government budgets has increased, and throughout the world healthcare systems are under budget pressure and actively managing their costs. With such strong headwinds, a sustainable future for MedTechs might be to reduce their emphasis on manufactured products distributed through labour intensive sales channels and increase their AI service offerings using data from their existing devices. Over the past five years AI solutions have become more prolific, easier to deploy, and increasingly sophisticated at doing what health professionals do, but more efficiently, more quickly and at a lower cost.  
 
~80% of B2B sales are digital

In addition to AI solutions being used to improve clinical outcomes, they can be employed to enhance business efficiencies. A previous Commentary described how AI systems can help to transform traditional labour intensive MedTech supply chains and personalise sales. A recent study undertaken by Gartner, a global research and advisory firm, suggests that, “Over the next five years, an exponential rise in digital interactions between buyers and suppliers will break traditional sales models, and by 2025, ~80% of B2B sales will occur in digital channels”. Giant tech companies are taking advantage of this to enter healthcare markets, MedTechs have been slow to implement such changes despite the boost in online engagements provided by the COVID-19 pandemic.
Reasons for slow adoption of AI systems

So, why are MedTechs slow to implement AI solutions to enhance clinical outcomes and improve economic efficiencies? Over ~3 decades they have achieved double-digit revenue growth from manufacturing physical devices and marketing them through labour intensive channels in a few wealthy regions of the world with relatively benign reimbursement policies. During this period of rapid growth and commercial success, MedTechs have not been required to confront data issues, bridge the science, technology, engineering, and mathematics (STEM) skills gap, and commit to new structures, new processes, new behaviours, and new aptitudes.
This suggests that despite a wind of change, now blowing through MedTech markets and challenging traditional business models and strategies, it could be perceived as a 'temporary breeze' and nothing will change. However, a step change in the direction of more AI solutions might occur when digital natives [people who have grown up in a digital age] replace digital immigrants [people whose careers were well underway before the onset of the digital age] in MedTechs’ C suites. According to a Gartner executive, “As baby boomers retire and millennials mature into key decision-making positions, a digital-first buying posture will become the norm. . . . . . Sales reps will need to embrace new tools and channels, as well as a new manner of engaging customers, matching their sales activity to their customers’ buying practices and information collecting needs”. A 2019 research report from the Boston Consulting Group (BCG), suggests that companies, which use AI systems to personalise sales can expect productivity gains of ~10%, and incremental revenue growth of ~10%.
 
Partnering with tech savvy start ups

Currently, many MedTechs neither have the mindsets nor the in-house STEM capabilities to create AI enhanced services. So, what might be a way forward? STEM skills, although scarce, tend to reside in people <30. Although there are ~68m of these people in the US, people with STEM skills tend to prefer to work either for giant tech companies or tech start-ups devoted to leveraging the potential of AI. Giant tech companies and start-ups are outside the comfort zones of most MedTechs. However, in the future, they may be obliged to partner with tech savvy start-ups engaged in developing AI driven solutions. Such collaboration will be challenging because it requires MedTechs to change their business models, create new ways of making strategic investments, and develop novel approaches to R&D that encompass a broader spectrum of partners.

Most of MedTechs’ R&D investment is consumed by incremental innovations to their current suite of devices. This tends to reinforce existing revenues rather than develop disruptive technologies aimed at capturing new revenue streams. Such strategies are efficacious in stable, fast growing economic environments, but lose their edge in slower markets. It seems reasonable to assume that, as market conditions tighten, MedTechs will need to consider shifting their R&D strategies towards the development of more disruptive technologies. We see this already in Stryker’s R&D investment in robotic surgical systems and Zimmer’s proposed R&D spend on AI, data informatics and robotics.

You might also like: 

China’s rising MedTech industry and the dilemma facing Western companies


and

Can Western companies engage with and benefit from China?
US supremacy challenged  

US tech giants are investing heavily in AI R&D and driving the adoption of advanced technologies in healthcare. Although these companies have made, and will continue to make, a significant contribution to the field, it would be a mistake to think that they have AI healthcare markets sewn up.
 
Three Chinese tech giants, collectively referred to as ‘BAT’, are also investing heavily in AI systems. All three offer services well beyond their core products and have far-reaching global ambitions. BAT is comprised of Baidu, China’s largest search provider, Alibaba the nation’s biggest eCommerce platform and Tencent, which runs WeChat that has access to >1bn users on its platform. For the past five years BAT has been expanding into other Asian countries, recruiting US talent, investing in US AI start-ups, and forming global partnerships to advance their AI ambitions.
In addition to these private endeavours, China has made AI a national project. Since 2017, Beijing has been pursuing a three-step New Generation AI Development Plan, which aims to turn AI into a core national industry. To this end, China is vigorously carrying out research on brain science, brain computing, quantum information and quantum computing, intelligent manufacturing, robotics, and big data. Already, China has become a world leader in AI publications and patents. The nation’s global share of AI research papers increased from 1,086 (4.26%) in 1997 to 37,343 (27.68%) in 2017, surpassing any other country, including the US. Most AI patents are registered by companies in the US and Japan. However, when it comes to AI patents registered by research institutes, China is the undisputed leader. According to a 2021 report on China's AI development,  ~390,000 AI patent applications were filed in China over the past decade, accounting for ~75% of the world total. Beijing’s competitive advantage in big data and AI strategies is driven by a combination of its weak privacy laws, a national plan, huge government investments, concerted data-gathering, and big data analytics by the BAT tech giants and others. Currently, China’s AI market is valued at ~US$22bn, and by 2030, the nation is expected to become a leader in AI-empowered healthcare businesses and the world’s leading AI power.

Beijing’s policies have given rise to hundreds of AI driven start-ups aimed at gaining access to new revenue streams in China’s rapidly growing healthcare market. Western MedTechs might consider accepting Beijing’s  Made in China 2025 policy, partner with these  tech savvy start-ups and jointly benefit from the nation’s current 5-year economic plan aimed at a “healthier China”.
 
Takeaways
 
We have presented an AI-driven prescription for MedTechs to enhance the quality of care while providing access to new revenue streams. We suggest that this can be achieved by bolting on AI solutions to existing devices, and over time through partnerships with tech savvy start-ups. But ~30 years of double-digit growth derived from manufacturing physical products and distributing them through labour intensive sales channels might have cemented mindsets among C suite incumbents that find it challenging to think beyond physical product offerings. This could suggest that the wind of change, now blowing through MedTech markets, will be perceived as a temporary breeze that does not require thinking beyond physical products, and AI solutions will be a long time coming.
view in full page
joined 10 years, 5 months ago

Yike Guo

Professor of Computing Science
Directory:
Health Policy
Expertise:
Bioinformatics
computing science
data mining
data science
EUTROC
knowledge discovery
large-scale data management
scientific data analysis
translational medicine

Yike Guo is a Professor of Computing Science in the Department of Computing at Imperial College London. He leads the Discovery Science Group in the department, as well as being the founding Director of the Data Science Institute at Imperial College.

Professor Guo also holds the position of CTO of the tranSMART Foundation, a global open source community using and developing data sharing and analytics technology for translational medicine.

Professor Guo received a first-class honours degree in Computing Science from Tsinghua University, China, in 1985 and received his PhD in Computational Logic from Imperial College in 1993 under the supervision of Professor John Darlington.

He founded InforSense, a healthcare intelligence company, and served as CEO for several years before the company's merger with IDBS, a global advanced R&D software provider, in 2009. He has been working on technology and platforms for scientific data analysis since the mid-1990s, where his research focuses on knowledge discovery, data mining and large-scale data management.

He has contributed to numerous major research projects including: the UK EPSRC platform project, Discovery Net; the Wellcome Trust-funded Biological Atlas of Insulin Resistance (BAIR); and the European Commission U-BIOPRED project. He is currently the Principal Investigator of the European Innovative Medicines Initiative (IMI) eTRIKS project, a €23M project that is building a cloud-based informatics platform, in which tranSMART is a core component for clinico-genomic medical research, and co-Investigator of Digital City Exchange, a £5.9M research programme exploring ways to digitally link utilities and services within smart cities.

Professor Guo has published over 200 articles, papers and reports. Projects he has contributed to have been internationally recognised, including winning the “Most Innovative Data Intensive Application Award” at the Supercomputing 2002 conference for Discovery Net, and the Bio-IT World "Best Practices Award" for U-BIOPRED in 2014. He is a Senior Member of the IEEE and is a Fellow of the British Computer Society.


view this profile