Algorithms help diagnose and treat diseases, discover new drugs and personalize prevention. Thus, the potential of artificial intelligence has been recognized not only by the medtech industry but also by health systems, pharmaceutical companies, patients and even the World Health Organization. What’s the state of AI in healthcare?
- PROBLEM: Gaps in data/information access have many negative consequences: medical errors, delayed diagnoses, rising healthcare costs, low quality of care;
- CHALLENGE: In order to fix broken healthcare systems, smart integration of new technologies into the current health ecosystem is needed.
- SOLUTION: AI can support this new paradigm shift by accelerating research, reducing the administrative burden on doctors, improving policy-making and enhancing health literacy;
- IN PRACTICE: Developing trustworthy AI/ML algorithms that meet the needs of patients/healthcare workers requires close collaboration between science and business as well as adequate funding.
In 2009, Hiroshi Kobayashi, a scientist from the Tokyo University of Science, presented the world’s first robotic teacher Saya. According to Kobayashi, machines are better than human teachers. The AI-powered robot knows the answers to all questions – it monitors and analyzes childs’ behavior to individualize the learning process and support their hidden talents.
This utopian vision of education has a lot to do with healthcare. Of course, nobody would like to be treated by artificial intelligence wearing a doctor’s coat – like no parent would prefer a robot over a human teacher. Human social empathy cannot be replaced by a rational machine. However, a lack of data causes medical errors, delays diagnosis and worsens the treatment prognosis. Humans make irrational health decisions, getting lost in conflicting recommendations and random advice.
AI’s capabilities to memorize the medical records of millions of patients, analyze data, monitor, forecast and make decisions based on facts are now becoming a critical transformational power in medicine.
Impact of AI in healthcare. New challenges require contemporary approaches
Doctors are drowning in administrative tasks, and the patient’s journey through the system is like a labyrinth. Healthcare professionals waste up to 1/3 of their time on administrative tasks. While WHO forecasts a shortage of 9.9 million doctors, nurses and midwives by 2030, such a waste of medical staff resources cannot be afforded.
There are many more statistics showing that healthcare needs radical changes: approx. 250,000-440,000 people die each year in the US due to medical errors. This is the third cause of death, right after cancer and cardiovascular disease. Cancer treatment delayed by one month increases the risk of death by approx. 10%. Non-communicable diseases generate 80% of healthcare costs. Paradoxically, 80% of heart conditions, heart attacks and diabetes can be prevented by modifications in lifestyle.
AI is a necessity, not just an addition to healthcare. Imagine the scale of the benefits when algorithms start reducing the number of medical errors by analyzing patient data and comparing treatment scenarios with the outcomes of millions of other people, relieving doctors of paperwork by automatically sorting information and collecting data in electronic records with voice processing systems.
“AI will bring the most significant benefit to patients and doctors by becoming the new member of the medical team. Automated solutions will assist physicians in medical decision-making, interpret radiology images and treatment plans, and even take over repetitive tasks. AI will help analyze a vast amount of patient data collected by health sensors, wearables and direct-to-consumer tests. In the future, patients will provide insights from AI-supported systems to their doctors. It will further facilitate their relationship with caregivers.”
How medicine, life sciences, healthcare professionals and patients benefit from AI
Released in September 2021 by the US Food & Drug Administration (FDA), the list of AI/ML-based (artificial intelligence/machine learning) medical devices already includes 343 items. By 2016, there were 15 items on the list. The largest number of medical algorithms certified by the FDA in the USA and CE-marked in Europe go to radiology. It is followed by cardiology, hematology, and neurology.
The advancement of digital and AI technologies in healthcare has led to a rapid increase in research in the field of AI and ML. So much so that in 2019, the prestigious scientific magazine The Lancet decided to release a separate version devoted solely to digitization – The Lancet Digital Health. Scientists from around the world publish research on (among others) the effectiveness of algorithms in healthcare (e.g., the one from April 2022 confirming the reduction in the incidence of colorectal cancer in the case of colonoscopy using AI tools.
Such a rapid development of algorithms is a logical consequence of the digitization of healthcare, cultural changes, and favorable legislative solutions. Systematically developed IT architecture in healthcare facilities and digitization of patient files facilitate the exchange and re-use of data by research centers and commercial companies.
Yet, it is no longer about advances in medicine – the competitiveness of the future economies will be measured by the scale of innovation and digitization maturity. The European Commission is aware of this and, at the beginning of May 2022, presented the European Health Data Space project. EHDS is intended to facilitate the secondary use of data, including supporting research and health policy goals.
The potential of AI in life sciences is being recognized by the pharmaceutical industry. Pharma leaders are working with startups developing so-called ‘digital therapeutics’ – platforms and mobile applications supporting patients in managing chronic diseases.
The range of potential applications is much broad. Novartis has partnered with a Chinese technology giant to develop AI Nurse – an intelligent platform that supports patients, doctors and nurses in managing heart disease. The program covered 500 hospitals. BioNTech, known for developing an mRNA vaccine for COVID-19, recently announced a multi-year partnership with InstaDeep Ltd., aiming to apply the latest advances in artificial intelligence and machine learning technologies to develop new immunotherapies for cancer and infectious diseases. And recently, Fujitsu has begun research to develop AI for early pancreatic cancer detection in Japan.
Big tech companies are also eager to enter the healthcare market. Alphabet, Google’s parent company, announced the launch of Isomorphic Laboratories in November 2021. The company aims to introduce an “AI-driven approach” to bio-pharmacy research by becoming a commercial partner for drug manufacturers.
Last but not least, public health has started to explore the power of AI. As a result of the COVID-19 pandemic, the WHO created the Hub for Pandemic and Epidemic Intelligence. By using AI to analyze health data, the WHO wants to prevent and limit future pandemics.
The gradual adaptation of AI by health and pharma organizations is an opportunity for digital health startups. Healthcare facilities are increasingly open to co-developing and implementing innovative solutions. Patients are also eager to use modern mobile apps to manage chronic diseases or optimize their lifestyles to stay healthy.
“There is space for improvement in what AI-enabled clinical decision support systems can provide to doctors for an early and correct diagnosis of diseases with overlapping symptoms. Tools that will succeed in integrating real-world data with clinical parameters and molecular profiles, leveraging AI for multi-dimensional data analysis, might significantly impact the definition of personalized medicine.”
Black boxes and ethical issues are among the biggest risks of AI in healthcare
Before the technological revolution gains momentum, some AI-related challenges must be tackled.
One of them is a low quality and limited availability of data used to train the algorithms. There are legitimate concerns about bias in data sets, which could lead to algorithms failing in populations not represented in the practice data—the so-called ‘black box’ (i.e., the way algorithms make decisions). Nobody knows how AI makes decisions, so it isn’t easy to verify the correctness of the entire process.
Citizens have concerns about data security and privacy. The use of diagnostic algorithms that are cheaper than human labor can potentially lead to two-speed healthcare, where diagnostic services provided by bots will be the standard, while contact with a human doctor will become a premium option, available beyond standard health insurance.
Doctors look at AI-based health systems with hope but fear. There are ethical dilemmas regarding professional liability for medical errors caused by algorithms. A mistake by an AI system suggesting the purchase of new clothes, a book or displaying the most exciting content on social media has no such consequences as an incorrect diagnosis or imprecise selection of a drug. Imagine such a bug in mobile apps used by millions of patients.
The legislation does not keep up with the rapid development of technology – many potentially beneficial solutions for patients are not scaled up on the market because health insurers do not reimburse them. However, this is also changing. In 2019, Germany adopted a new law that allows doctors to prescribe certified apps to their patients. Now France will implement a similar legislative framework.
The adoption of new technologies also requires a change in the work culture in healthcare. The case of IBM’s Watson Health proves that we are still far from balanced cooperation between doctors and AI. Neither the technology nor the doctors were ready. As a result, Watson Health was accused of making inaccurate and unsafe recommendations, prompting many hospitals to sever their cooperation with Watson. IBM eventually sold Watson Health.
The future of AI: quantum computing will accelerate big data science
AI advances will be strongly related to advances in hardware. Solutions enabling the processing of large data sets and the detection of correlations in data invisible to the human eye (or rather, traditional statistics) have only become popular in the last decade. We are talking about so-called ‘deep learning/machine learning’ (DL/ML). And although the history of neural networks dates back to 1943, DL/AM entered practical use in this century. AI needs not only data, but also computing power. And this – according to Moore’s Law – grows exponentially: the number of transistors on a single microchip doubles every two years.
Yet, another innovation will fuel the development of AI – quantum computers with computational power incomparably higher than traditional computers. Google says its laboratory version of a quantum computer is 100 million times faster than any classic computer. Its performance can be compared to the strength of 5 million laptops. In tests, Google’s 54-qubit computer was able to complete a task in 200 seconds that would take over 10,000 years on traditional computers. Such machines can speed up, hundreds or thousands of times, the time it takes for AI systems to search for new molecules for potential drugs.
In 2017, a Chinese AI-based robot passed a medical exam. But even in China, with ambitions to become an AI leader by 2030, robots have yet to replace doctors. Instead, they support them. And one group will benefit most: patients.
The link between AI, healthcare transformation and innovation ecosystems
Advances in healthcare and life sciences need a favorable ecosystem that fosters collaboration and supports the most brilliant ideas of young entrepreneurs. One place where these growth drivers can be found in Europe’s leading healthcare and life sciences hub is the Basel Area – a hotspot for progress in life sciences and medicine.
The people, culture of inventiveness, and location make it an excellent place to co-shape the future of healthcare. Based in Basel (Switzerland), it’s within walking distance from leading pharma companies working on breakthroughs in life sciences and – for a few years – also applying AI across their value chain. Here, Novartis cooperates with Microsoft within AI Innovation Lab, IBM plans a new center for quantum computing, while numerous international startups find symbiosis with the local supportive environment for entrepreneurship.
Based on this ecosystem, the DayOne is an accelerator that focuses on creating the ideal conditions for disruptive innovation in the life science industry. Recently ranked in the top 10 accelerators in Europe for health technologies, the DayOne initiative offers a platform for collaborative innovation across disciplines and industries by accelerating startups, launching catalyst projects, and hosting a community and events to bring innovative healthcare ventures forward. Since 2018, the hub has supported 23 companies with a combined valuation of over CHF 80 million, contributing to the life sciences ecosystem of the Basel Area, helping to launch, connect, and mentor many innovative healthcare ventures in medtech, medical devices, digital health, AI and more.
One example is Zoundream which uses AI to monitor baby cries to identify infants’ needs, emotions, well-being, and physical and neurological status. Another of its companies, Rekonas, has developed AI for EGG analysis to assess brain health, current, future cognition, and sleep macro/micro-structure. Or Nutrix, which applies AI to analyze molecules and biomarkers in saliva samples for health monitoring.
The future, AI-driven healthcare starts from ideas that can grow in the right ecosystem. The DayOne initiative represents a great example of the need to invest in ecosystem activators to achieve healthcare transformation.