DMEA: How AI is being used in medical practices and clinics

Various start-ups presented their products at the DMEA. The AI models Argo and Aureon from IDM gGmbH are designed to help with documentation in hospitals and relieve the burden on doctors. IDM is a non-profit limited company and a subsidiary of the University Medical Center Hamburg-Eppendorf (UKE). While Argo is intended to help with the creation of doctor's letters, Aureon is a text-to-speech model that has been specially trained for medical terminology and can be used in various areas such as nursing. A Large Language Model (LLM) was trained with data from the UKE to create the doctor's letters.

In his presentation, IDM CEO Nils Schweingruber pointed out that AI systems are often paid for twice – with money on the one hand and with data on the other: “Most of this data and the money ends up with big tech companies outside Europe, and they are of course happy because they get a lot of money and our health data […] We actually want an AI that meets our data protection standards, our security standards, but very importantly, also our quality standards”. This is particularly important in a medical context. He is consistently pursuing his vision at the UKE, where a completely digital patient file has been kept since 2009. The model accesses the entire documentation in the patient file and generates an epicrisis – i.e., a final report from a treating physician.

Since January of this year, Aureon, which is hosted in Germany, has been available to all 15,000 UKE employees and is also being used successfully by nursing staff and administration. “We will make Aureon partly open source,” announces Schweingruber, to make access to this technology even more widespread. “We simply want no one in the German-speaking healthcare sector to have to do without speech-to-text anymore.” Aureon will be available as a cloud solution for the outpatient sector this month, “so that it can be used independently of a local hardware infrastructure”. According to the press release, the UKE has also been using the transcription tool Orpheus since the beginning of the year.

Docport wants to digitize practices

Docport GmbH also gave a further insight into the digitization of medical practices – from medical technology to practice management systems with a subscription model. Mikail Bahar, who is a doctor and co-founder of Docport, explained that the sovereignty over the processed data remains with the medical practices. The company relies on federated learning so that the data is not sent to the AI cloud, “but the AI models are brought to the data” – into the “territory” of the medical practices. This now results in millions of data points per year. Bahar sees the combination of data analytics and optimized processes as a lifesaver.

(mack)