Should I turn to machine learning?

Artificial intelligence How machine learning can support medicine

In the future, pattern recognition should detect diseases much earlier than the general practitioner, and genome analyzes to identify genetically caused diseases should become more precise. But the AI ​​also inspires scientists who think in a completely different direction, explained Dlf IT expert Maximilian Schönherr. For example, in the care of patients at home or in the home. For example, a Berlin start-up is currently working on an app that determines a value that plays a major role in elderly care: the index for mobility. It must always be reported to the medical service of the health insurance companies up-to-date.

Collect, unite, extrapolate data sets

Siemens Healthcare GmbH, on the other hand, is working on a different question: How can you capture the mass of patient data for nursing staff in such a way that work can be simplified and the urgency of patient concerns better assessed? The problem: demographic data is often available to the doctor, other medical data is sometimes distributed over several technical devices, and not all of them have internet access. Machine learning could help in the future that algorithms are scalable and generalizable and that they can also be used with more variables and more data. However, such AI programs and apps still have to be tested on real patient records. In general, the following applies in medicine: the expert and not the technology has the last word, explained Maximilian Schönherr in the Dlf.

Sensitive data and AI

In medicine in particular, a huge amount of sensitive data plays a decisive role. Therefore, data protection is extremely important. The IT in hospitals deals with the use of artificial intelligence, among other things, with the anonymization of data for research purposes.

In contrast to the previous conference, this time the amount of data that is needed for so-called deep learning also played a role. Large amounts of data are important in many areas, said Schönherr in the Dlf, if only to clarify whether an AI model works.