Artificial intelligence in healthcare: Karlsruhe’s IT competence ensures safe applications
The use of artificial intelligence (KI) in medicine promises major benefits for patients.
AI-based assistance systems help early identification of illnesses, enable rapid analysis of large volumes of image and laboratory data, and offer the opportunity for individualised therapies.
Prof. Jörn Müller-Quade of the KIT Karlsruhe Institute of Technology addresses the issue of better healthcare with the aid of AI and the ensuing challenges for IT security: “Health records are a huge treasure trove that must be learned from ... this is important, particularly in these time of a global pandemic, in order to recognise patterns and develop strategies. The bigger the volumes of data are, the greater are the insights gained from them. The challenge is to use the confidential, personal data and at the same time evaluate it without infringing privacy.”
Due to interprofessional collaboration in medicine, therapy, and care along with the involvement of other players in the healthcare sector, many people require potential access to patients’ records. “This makes it difficult to protect sensitive heath data from unauthorised access”, says the Karlsruhe cryptographer. “Information technology is a facilitator, but only with a very strict rulebook and multiple levels of security for safe data transmission and access control.” The Federal Ministry of Health is currently preparing the foundations for the electronic prescription and electronic patient file (ePA), “but if the terminal devices, i.e. the PCs of chemists, health insurance companies, and medical professionals, are not secure, we will be faced with problems that could lead to data scandals”, says Müller-Quade, who heads the cryptography and security research group at the KIT and is the initiator of the KASTEL centre of competence for IT security.
Learning systems must be monitored by people
In addition to manufacturers’ liability, he also considers certification of AI systems and databases in medicine as well as the electronic patient file to be an essential requirement. “Health records must not be permitted to be used to the detriment of patients, for example where someone could be refused employment due to the disclosure of pre-existing medical conditions.”
The planned introduction of the ePA will give patients full control over their health records, which are then stored on their own PC as well. This makes it all the more important to adhere to the basic principles of IT security – to always use the latest operating system and secure passwords –, otherwise the PC can become an open gateway for intruders and reveal a patient’s medical history.
The computer scientist sees the use of constantly-learning AI systems whose software changes when operating without human surveillance as “a mixed blessing. The decisions taken by an AI system of this kind must always be checked for plausibility by the people performing the treatment. “Doctors must not apply the suggested result without reflection.”
Further information on AI in medicine can also be found in the following whitepaper with a cancer patient as an example: