Photo: Pulsub/Getty Images
The Himss25 session will show how a immense language model can monitor the radiologist’s notes to protect patients against medical errors, and receive recommended follow -up visits.
Medical error, including improper or delayed diagnosis, is one of the most significant causes of death in the United States. Delayed or omitted diagnosis (mod) are particularly common in diagnostic imaging, in which random results require an additional assessment to end the assessment of potential pathology. According to Parkland Health, the main healthcare system, it is called that delayed imaging surveillance is the main healthcare system.
According to the Public Health System in Dallas, Texas, Parkland, 1.7% of all CT and MRI research includes such findings.
“This session will offer clear possibilities of this type of program in a high number, security, net healthcare in which resources can be limited,” said Dr. Treacher, senior scientist at PCCI, Parkland Center for Clinical Innovation.
Treacher and others from Parkland say during the Himss25 session “Creating a large language model for the catalog of important radiologist’s recommendations”, which take place on Wednesday, March 5, from 3:15 to 16:15 in Venetian | Level 5 | Palazzo o at Himss25 in Las Vegas.
Parkland researchers have developed a immense language model that identifies and flags delayed recommendations regarding supervision from interpretation of radiologists. LLM has been integrated with electronic health documentation in Parkland, enabling centralized management and navigation of these cases. The results show 95% accuracy in imaging identification, which requires observation based on a doctor’s notes and 85% accuracy in determining the appropriate observation date.
“The large language model exceeds a manual review, which can be burdensome, time-consuming and more susceptible to errors, and we have found an accuracy of 98.1% to detect LLM observations based on our experiment,” Treacher said.