Viktoria Steinbeck explores if the EU is ready for Artificial Intelligence (AI) based healthcare solutions

Is AI the solution health systems have been waiting for?

While artificial intelligence (AI) is often associated with robots with superior intelligence that will take over the world, AI is much more down to earth than some expect. Through the use of basic analytical functions, AI can facilitate processes and decision-making in different spheres of life. In the healthcare sector, making a correct and timely decision can be a question of life or death, which is why AI is often praised as holding the potential to revolutionise healthcare systems as we know them.

How can AI help address modern day healthcare challenges?

Medical advances to date have greatly extended life expectancy across the world. Yet significant challenges still exist, for example in the management of chronic diseases, in oncology or in the sustainability of healthcare systems. These challenges require smart solutions, decision-making support and the automation of tasks to free up resources for innovation and time for physicians to do what they do best: care for patients.

Among other aspects, AI can help improve diagnostic accuracy, support precision medicine and accelerate research, and thus innovation. AI solutions could also help to create more sustainable – and more equitable – healthcare systems by transforming and optimising care delivery.

Probably the most well-known application of AI in healthcare is enhancing diagnostic accuracy in imaging, such as in pathology, dermatology and radiology. Deep learning algorithms have already been shown to make diagnoses as accurately as physicians. For instance, according to a study published in 2016, in combination with physicians’ diagnostic abilities, deep learning predictions have been shown to increase diagnostic success rate to 99.5% – equalling an 85% reduction in human error rate!

Their ability to recognise patterns in data that may have previously been unrecognisable or too difficult to detect makes algorithms the perfect tools to support individualised care as well as improve and speed up much needed innovation in the health sector. It is not by chance that more and more pharmaceutical companies are partnering with AI start-ups. And as Cochrane’s “Transform” project in collaboration with Microsoft shows, AI is a great way of accelerating the generation of systematic literature reviews, making it easier for medical professionals to keep up to date with the latest developments.

Everyone talks about data safety… but what are other potential risks?

Data safety is probably the most widely discussed risk related to AI. But it is only one among many. AI applications come with various risks associated with the way they learn and make decisions. While algorithms generally perform well with a large amount of high-quality data, they are only as good as the data they are fed. This can lead to underperformance, for example, in cases of novel drugs’ side effects where no prior information was recorded. Deep learning also holds the risk of turning computers into “black boxes”, making it almost impossible to understand how exactly they came up with a decision and potentially leading to biased or unethical medical recommendations.

AI can also significantly impact on the doctor-patient relationship. Doctor-patient interaction relies to a great extent on non-verbal cues, which are culturally coded and could be easily missed by AI technologies. Who is to be held accountable when something goes wrong?

The need for concrete guidelines and robust trials is clear. At the same time, AI solutions are much needed and should be brought to the market quickly so that patients can benefit from the next wave of medical innovation.

So, what is the EU doing about it?

With the European Commission’s President-elect promising to create an AI plan during her first 100 days in office, the topic is clearly gaining recognition at EU level. However, as Ursula von der Leyen is still waiting for her Commission to officially take office, this timeline might not be achievable.

But there is movement in Europe – even in Germany, a country where data protection is very high on the agenda. With the dedicated German data ethics commission having recently proposed their AI strategy to Angela Merkel, calls for an overarching EU law on AI similar to the GDPR have already been heard. Will the German proposal create the basis for the development of AI laws in the EU?

The EU’s AI experts, and AI companies might not agree with this plan as they have been advising against strict horizontal rules, which it is feared would stifle innovation as well as the speed and possibility of adapting new AI innovation across sectors.

But the Germans seem determined and with AI being one of the core focus areas of the upcoming German presidency of the Council of the EU and the German Digital Health Law being passed just last week, it all seems to point to a decided German push for action on a new EU framework for AI effecting healthcare innovation across the EU.

Regulation that ensures safety of data and AI based decision-making is paramount but needs to be practical. If patients are to benefit from AI healthcare solutions, the EU and Member States need to step up their game to provide incentives, education and market access programs that accelerate and incentivise high quality AI in the EU.