AI used to better understand the mechanism of trauma and improve its management. The solution is proposed by researchers from Inserm and the University of Bordeaux. They have designed an algorithm to analyze reports about patients admitted to the emergency room, says an Inserm press release, published on May 4.

The researchers started from an observation: trauma concerns a third of emergency room care. This represents 21 million hospital visits each year, whether for everyday or road accidents, violence or suicide attempts. As soon as a patient is admitted for a trauma, caregivers fill out reports that detail the symptoms, the patient's condition and the circumstances of the trauma, says BFMTV. All this in medical jargon but also with common terms, abbreviations, without it being perfectly structured. "These data remain unexploited today," says Inserm.

#IA: Researchers have developed an algorithm capable of classifying visits to #urgences (97% accuracy) due to #traumatisme thanks to the analysis of clinical reports by an artificial intelligence #GPT.

Press release:

— Inserm (@Inserm) May 4, 2023

Access to this content has been blocked in order to respect your choice of consent

By clicking on "I ACCEPT", you accept the deposit of cookies by external services and will thus have access to the content of our partners


And to better pay 20 Minutes, do not hesitate to accept all cookies, even for one day only, via our button "I accept for today" in the banner below.

More information on the Cookie Policy page.

500,000 reports analyzed and 97% success rate

The idea is therefore to put things in order, thanks to artificial neural networks of the GPT type. The tool developed by the researchers was called Tarpon (for Automatic processing of emergency room summaries for a national observatory). The AI was trained using 500,000 reports from the adult emergency department of the Bordeaux University Hospital. The algorithm was able to automatically classify traumas according to their types, "and this with surprising precision". It appears that 97% of these reports were filed correctly by the tool, compared to 86% with the old methods, which are probably more time-consuming.

The objective of this tool is "the establishment of a national trauma surveillance system, but also epidemiological analyses concerning, for example, the impact of drug consumption on the risk of accident", details Inserm. After Bordeaux, the project will be extended to about fifteen emergency services in the territory.

  • Tech
  • Artificial Intelligence (AI)
  • Health
  • Emergency room
  • Inserm
  • Hospital