Algorithms in medicine: AWMF working group of doctors and lawyers discusses the use of artificial intelligence in medicine and the resulting liability rules

09.06.2022 16:16

Whether in diagnostics, therapy or research – the use of artificial intelligence can help to improve medical treatment in everyday life. In its most recent meeting, the working group of doctors and lawyers of the Association of Scientific Medical Societies eV (AWMF) discussed what conditions are needed so that doctors can make better decisions with the help of algorithms. From the perspective of the panel of experts, one of the things that will be important for the future is to clarify open liability issues.

The possible uses of artificial intelligence (AI) in medicine are diverse. It can help doctors make the correct diagnosis, for example by training algorithms to recognize certain diseases in computer tomography scans. Digital applications in the form of apps are already available for patients, which show causes and possible solutions based on the symptoms and findings entered.

The use of artificial intelligence also offers opportunities for medical research: “By using algorithms, various research data sets can be networked with each other and analyzed at high speed. This could help to achieve new scientific findings much more quickly in the future. “Today it takes far too long for results based on a research hypothesis to result in changed guideline recommendations,” says Professor Dr. rer. nat. Martin Sedlmayr, holder of the professorship for medical informatics at the Carl Gustav Carus Medical Faculty at the Technical University of Dresden.

“The areas of application of AI in medicine are diverse. It is clear that AI cannot replace the doctor. Rather, it should support those treating them in decision-making and be used where it benefits the patient,” says Sedlmayr. Against this background, the expert believes it is necessary for the field of AI to be more closely integrated into the training of students - both in computer science and medicine.

The prerequisite for the successful use of algorithms in medicine is data contained in the system - sufficient and of appropriate quality. “The availability and quality of the data from which algorithms are fed determine the quality of the results of AI applications,” emphasizes Professor Dr. med. Alexander Cavallaro from the Imaging Science Institute (ISI) in radiology at the University Hospital Erlangen. Today, data is usually available in different forms, for example as text documents in a patient file or as images such as x-rays.

“Before an algorithm can be applied, the data must first be extracted and organized. The aim is, for example, to prepare different content such as texts in findings, images or laboratory values in such a way that, when networked with one another, they can generate appropriate solutions,” explains the expert. The AWMF has repeatedly called for quality criteria: transparency and trustworthiness of the data basis. To do this, knowledge from high-quality, evidence-based guidelines must be integrated into rule-based algorithms and AI applications as the primary database and made visible to users.

Overall, doctors and lawyers agree that the use of artificial intelligence will play an important role in medicine in the future. “It is crucial that humans remain in control of the AI. It is important that doctors are always able to identify possible errors and their sources. This in turn requires that they can understand and comprehend how artificial intelligence is created and how the algorithm used achieves results,” emphasizes Cavallaro.

The AWMF working group of doctors and lawyers was therefore also concerned with clarifying open liability issues at their conference. The legal scholar Professor Dr. iur. Oliver Brand explained: “Doctors must always remain responsible for monitoring the results of artificial intelligence. This means they continue to be responsible for diagnoses or therapeutic options, even if they were determined with the help of AI.” However, with self-learning systems, attributing the effects of AI to doctors causes problems. “Because the civil law institutes when it comes to the question of attributing liability to their

Limits are reached, new legal regulations are needed. Until then, the current rules represent a useful emergency solution,” says Brand, holder of the chair for civil law, private insurance law, business law and comparative law at the University of Mannheim.