Icon Berlin Health Data Space – AI to beat acute kidney failure

Practical example and current challenges

  • Acute kidney failure is a frequently unrecognized problem in hospitals. Up to 20% of all patients suffer acute kidney failure during an inpatient stay, which leads to a restriction of kidney function within a few hours or days.
  • In the early stages of acute kidney failure, the disease is often not detected, although effective treatment options are available. Mortality rates of acute kidney failure are about 20% and can rise to over 50% in severe cases. The lack of cross-hospital data makes it difficult to establish an algorithm for early detection and thus the timely diagnosis of acute renal failure.
  • The Use Case aims to establish algorithms for the early detection of acute kidney failure and to detect acute kidney failure earlier by exchanging data (e.g. Laboratory values, diagnoses, procedures) between Berlin hospitals. A platform for the exchange of data and algorithms between hospitals in Berlin and an AI-based alarm system will be established as a prototype.
Berlin Health Data Space – AI to beat acute kidney failure

What added value does the "GAIA-X project" offer?

  • The creation of a distributed data room, based on GAIA-X for the exchange of data and algorithms between Berlin hospitals, together with an AI-based alarm system, can improve early detection and, thus, reduce mortality.
  • Already on arrival at the hospital, patients benefit from better data exchange and the development of predictive models for early detection.
  • In addition, GAIA-X enables cross-hospital networking and fulfils the basic requirements for providing hospital-typical algorithms by means of distributed learning and for supporting medical staff at an early stage of treatment.
  • The project supports the use of a secure infrastructure with a decentral organised database and, thus, contributes to data sovereignty. Only necessary health data is queried, and the algorithms are moved to the data.
  • Through methodical application of distributed learning, sensitive health data can be processed locally in accordance with the law, e.g. directly in the hospital. A secure, networked and multifunctional cloud environment also enables the use of the latest AI analysis methods.

Use Case Team

  • Prof. Dr. Klemens Budde – Charité – University Medicine Berlin and Plattform Lernende Systeme
  • Dr. Thomas Schmidt – acatech and Plattform Lernende Systeme
  • Dr. Johannes Winter – acatech und Plattform Lernende Systeme