Software and security specialist – Why the digital revolution in medicine still has a long way to go

If many people's wishes are fulfilled, the digitization of the healthcare system in Germany would have to pick up significantly. According to a recent representative survey by the digital association Bitkom, 71 percent of people are in favor of accelerated digitization in this area. At the same time, however, around half of those surveyed feel overwhelmed by the ongoing changes. To counteract this, Bitkom Vice President Christina Raab demands: “We must strengthen the skills needed to deal with digital health technologies and applications.”

That is absolutely true in two respects: both people themselves as patients need to acquire new skills for using digital technologies and those who work in the healthcare sector. This aspect is becoming increasingly important as digital transformation progresses. And not only that: developments in the field of artificial intelligence in particular have the potential to turn a lot of things in the healthcare sector on their head.

Peter Liggesmeyer is the director of the Fraunhofer Institute for Experimental Software Engineering IESE in Kaiserslautern and holds the chair for software engineering in the Department of Computer Science at the RPTU Kaiserslautern-Landau. In addition to the focus topics of artificial intelligence, autonomous systems and Industry 4.0, his research focuses primarily on concepts in the area of ​​safety and security.

Many possible applications of AI

This doesn't even have to be about how AI-supported robots will be used in operations in the future. The possible uses of AI are already much more low-threshold. Let's take the issue of bureaucracy, for example. In the healthcare sector, there are still a lot of manual documentation processes. Doctors have to spend a lot of time documenting their work and putting it down in writing. This time is missing for other activities, such as having a detailed conversation with patients.

But that could soon change. With the help of AI, these documentation processes could be carried out almost automatically. This in turn would benefit everyone involved: the patients, the nurses and the doctors.

The situation is similar with the questionnaires that patients have to fill out in the waiting room. Firstly, these are often pages of text that are not really prepared in a user-friendly way. Secondly, patients usually have no opportunity to ask questions in peace. And thirdly, in the worst case, the questionnaires end up in a drawer somewhere without being given any further attention.

This doesn't have to stay that way in the future either. With AI, this process could be significantly optimized, for example by making the questionnaires available digitally and having integrated chatbots answer direct queries in a standardized manner. The questionnaires would then be saved digitally so that doctors can access them at any time, even later. Or what if an AI compressed the latest medical research reports into the essential core aspects so that doctors can keep up to date easily and simply?

Insufficient data available

I could give many more examples, and some AI-based systems are already being used in the healthcare sector. Although the potential is really great, there are also a whole series of challenges that make the use of AI not quite as trivial as it seems at first glance. One of these is the very thin data situation that currently prevails in the healthcare sector.

AI systems depend on extensive data sources with high-quality data; at least if they are to function well and reliably. This data is often not available in sufficient quantity and quality, especially in medicine, or cannot be used without further ado. This makes it extremely difficult to develop new AI-based solutions for this industry.

It is particularly challenging to find a suitable balance between data protection and targeted data use. After all, the healthcare sector works with a lot of data, often very sensitive data. On the one hand, it is of course important to set high standards for data protection in this area. On the other hand, however, this must not lead to data no longer being used at all. Particularly in the context of medical research, the responsible use of data is essential in order to generate new findings and solutions for the benefit of patients.

Dependable AI as a major research field

The technical hurdles that still exist are probably among the biggest challenges. This mainly concerns the large research field surrounding Dependable AI – in other words, AI systems that are reliable in terms of safety and security. Even though the field of AI is much broader, the current discussion in this area essentially revolves around machine learning and neural networks. These are developed very differently than classic software, which causes advantages as well as disadvantages.

While in “normal” software the behavior is specified in a programming language, neural networks are trained with large amounts of data. This means that they “learn” their functionality from the examples contained in the training data. This can work well depending on the quality and amount of training data. However, there is no guarantee that the neural network trained in this way will always react as desired. It is therefore not yet possible to prove that such software always does what it is supposed to, which rules out these solutions for certain critical areas of application.

I deliberately say “still” here, because this is exactly what research is currently being carried out at full speed – including here at Fraunhofer IESE. If we succeed in guaranteeing this reliability for such AI systems, the application potential for the healthcare sector – but of course also for many other areas – seems almost unlimited. But there is still a long way to go before we get to that point. Until then, such systems will support the healthcare sector – no more, but also no less.