In healthcare, all eyes are on Artificial Intelligence (AI). New AI-driven technologies have the potential to boost personalised medicine, automate laborious tasks and facilitate data-driven clinical decision making. With every wave of new products and services being announced and launched, stakeholders’ expectations seem to reach new highs. However, only few solutions have received market authorisation, are reimbursed and/or have been widely implemented in clinical settings.
“We should stop trying to beat the radiologist but use the true potential of AI”
– JASPER LEVINK –
In our series of publications about AI we reflect on these developments to guide life sciences researchers, entrepreneurs and clinicians in their efforts to transform healthcare with AI. In this first article Jasper Levink, shows his vision on AI and how a solid business approach is key for success. “We should stop trying to beat the radiologist but use the true potential of AI”.
Vision of AI
How can AI be transformational in health care and life sciences? Jasper: “AI can be transformational in leveraging real-world data for e.g. clinical decision making. We still don’t understand much of the majority of diseases: how and why they occur, and why they manifest differently between one patient and another. Although the answers are encoded in the data trails all of us produce continuously, their magnitude make them difficult to comprehend for humans. AI can extract what we need, or at least guide us to the places where we can find the answers. The challenge is how to make these data streams computable and extract the right insights from them. But we need to be smart: AI without a vision is like a fishing expedition.”
Learning health systems
How can clinical information that is now difficult to comprehend for humans and hence underused, become useful using AI? “Currently there is only a limited amount of clinical data a doctor can oversee. On top of that, the process is highly inefficient. A clinician may be able to check and compare patients’ characteristics for the past months or some years, but not for the past decennia. Moreover, currently there is no effective way to incorporate real-world data. AI can open up clinical and real-world data by automatically identifying comparable cases and hence can be pivotal in the larger transformation from a protocol-driven to a learning data-driven health system.”
Beat the radiologist
Are the current AI solutions transformational yet? “Not really. Unfortunately, in both diagnostics and treatment, a lot of AI potential is left unused. There is too much focus on trying to ‘beat the radiologist’. Basically, we are trying to manufacture a car with a robot in the same way we used to make that car by hand. When we find a way to enclose patient outcomes in computable large-scale data sets, AI can move beyond automation and can become truly transformative.”
Explainable and trustworthy AI
But will clinicians accept such solutions? “That is an excellent point. In order to ensure that clinicians continue to feel in control, AI algorithms should remain explainable and trustworthy. This means that the underlying decision should be transparent and that the algorithm understands when it is failing and/or reaching its limits.”
Alignment of research and business creation processes
What is needed for a valuable AI solution to make it to the market? “We often see a misalignment between the research and the business creation process of AI. The time spent on validation, certification, reimbursement and marketing needed to successfully bring a product to market, drastically exceeds the time of developing an AI algorithm. The efforts required to ensure reimbursement and uptake are typically underestimated. Often, even before an algorithm has been put to work, newer and more competitive ones have been developed. This challenges the investment case of AI solutions. Hence, on the short-term, research efforts on AI solutions for healthcare should focus on those where validation and reimbursement are manageable, or hybrid business models can yield early revenues. On the longer term, regulatory frameworks for self-learning algorithms and/or rapid update cycles are needed to ensure also other AI applications represent viable investment cases.”
Read our whitepaper
In our whitepaper and blog series we give you our vision an AI. We discuss the basic workflow of AI tools, we provide an overview of the opportunities and challenges of implementing AI in healthcare and give actionable insights to help the different stakeholders in overcoming barriers and developing smarter strategies that accelerate the introduction of AI innovations.
Download whitepaper AI
In this whitepaper ttopstart will give you insights on accelerating Artificial Intelligence for healthcare and life sciences. This paper will provide an overview of the opportunities and challenges of implementing AI in healthcare and give actionable insights to help the different stakeholders to overcome barriers and develop smarter strategies that accelerate the introduction of such AI innovations.