Fracture detection AI model performs well in real-world setting

Monday, December 2 | 1:30 p.m.-1:40 p.m. | M6-SSMK03-1 | E450A 

If you're interested in hearing about the real-world performance of a fracture detection AI algorithm, you won't want to miss this update from a project at four hospitals in Norway.

In an effort to reduce emergency room wait times and doctor consults, the Vestre Viken Hospital Trust implemented a CE-marked AI application to analyze x-ray images of patients presenting with trauma. Pre-implementation, each hospital had various workflows for handling these patients until treatment.

Radiographers used the AI application's results as preliminary reports to navigate patients accordingly, with triaging incorporated within the radiology information system and a flagging system for the AI results, according to Vestre Viken implementation lead Line Tveiten and Ramprabananth Sivanandan, MD, who will present on behalf of the project. 

In advance, the team highlighted that the AI-enabled radiographers to directly discharge 4,697 patients with negative findings except in the case of clinician requests, thereby reducing total patient waiting time by 201.3 days in total on analysis of 20,083 patient examinations. The AI also decreased the need for 2,348.5 doctor consultations, Tveiten and Sivanandan noted in their abstract 

The team also collected data on workflow types, patient waiting times, necessity for doctor consultations, and the triaging process. Postimplementation, they consolidated six workflow pathways into a single standardized pattern which significantly improved the function of the radiology department and hospital efficiency and effectiveness across the four hospitals. 

Stop by to find out more about what else the research team measured as well as clinical AI implementation tips.

Page 1 of 2
Next Page