Opportunities and risks of artificial intelligence in medicine

Experts discuss the use of small amounts of data at workshop

Artificial intelligence is an integral part of many areas of everyday life. It can also help in medicine, for example to filter out correlations between symptoms and diseases from medical data and make treatment recommendations. But what happens when only small amounts of data are available? Can AI then still work at all? Professor Dr Klaus Eickel from Bremerhaven discussed this with thirty international scientists at the three-day scoping workshop ‘Data Augmentation and Imputation Methods for Health Data’ in Hanover. This workshop, which was funded by the Volkswagen Foundation, was initiated by a cross-institutional working group from the Bremen scientific community through activities of the U Bremen Research Alliance (UBRA).

Prof Dr Klaus Eickel conducts research at Fraunhofer MEVIS on the development of novel, non-invasive biomarkers for the diagnosis of Alzheimer's and MS. He teaches on the medical technology programme at Bremerhaven University of Applied Sciences as Professor of Medical Informatics. 

In order to further develop research areas, scientists need freedom away from their research projects. Scoping workshops organised by the Volkswagen Foundation offer this opportunity. Here, international experts jointly develop ideas on identified research gaps and engage in intensive dialogue. This is also a great opportunity for medical informatics. ‘When it comes to data in healthcare, the potential of artificial intelligence to support diagnoses, predict outcomes or make treatment recommendations seems very promising. However, a major challenge is that often only relatively small or incomplete amounts of data are available. One solution is data augmentation, where the AI automatically fills in the gaps,’ explains Prof Eickel. 

But of course it's not quite that simple. It must be ensured that the data generated by the AI does not falsify the result. ‘We have to define quality criteria and check the influence of the additional data on the result,’ says Prof Eickel. This requires expertise from various disciplines. The participants in the scoping workshop discussed what expectations should be placed on the AI application and which known methods can already be used. Together, they are now working on a position paper that will be published later this year. 

(Our press releases are translated with the support of deepl.com.)

Scientific contact person
Prof Dr Klaus Eickel
Professor of Medical Informatics
E-mail: keickel@hs-bremerhaven.de
Phone: +49 471 4823-942