by
Lauren Dubinsky, Senior Reporter | August 22, 2024
Additionally, the AI tool had equal or lower rates of critical misses than radiology reports at sensitivities of 95.4% or higher. But Plesner noted that the AI's mistakes were more clinically severe for patients than the radiologists' mistakes because as he mentioned earlier, the radiologists were interpreting findings based on the clinical scenario.
"I envision that [the AI tool] will be a triage tool initially, for sorting the worklist, but it won't initially be used for automatic purposes," Plesner explained. "After such a pilot, the system can move toward automatic functionality when there is trust in the system and one has collected some data on missed findings."

Ad Statistics
Times Displayed: 109402
Times Visited: 6640 MIT labs, experts in Multi-Vendor component level repair of: MRI Coils, RF amplifiers, Gradient Amplifiers Contrast Media Injectors. System repairs, sub-assembly repairs, component level repairs, refurbish/calibrate. info@mitlabsusa.com/+1 (305) 470-8013
He then suggested employing it at a high-sensitivity threshold while still keeping an eye on performance. From there, the sensitivity can gradually be decreased to 98%.
"However, if there is an institution where chest X-ray is not reported at all per standard of care, then I think this could be deployed straight away using a disclaimer in reports that it was AI generated," he concluded.
Back to HCB News