Monday, November 30 | 8:00 a.m.-9:00 a.m. | M1-SSGU02-8 | Room E353B
In this talk, researchers will make the case that AI can help radiologists read prostate MRI exams more accurately and consistently.
Prostate MRI interpretation varies with patient population, imaging protocol, and reader experience, which leads to inconsistent cancer detection, localization, and biopsy decisions, according to presenter Brittaney Everitt of DeepHealth, and colleagues. The group sought to determine if an AI platform that offers automated lesion segmentation and lesion-level risk classification could improve detection and reader consistency.
In a blinded, multi-reader study, 12 board-certified radiologists interpreted 247 prostate MRI exams twice, once with AI and once without, separated by a washout period.
With AI, reader cancer detection rose by nearly 16%, and every reader improved. The gains held across subgroups, including patient age, race, and ethnicity, scanner vendor, field strength, and reader experience, the researchers observed. Agreement between readers more than doubled, suggesting they often accepted the AI-generated segmentations with minimal changes, according to the researchers.
“Generalizable across diverse populations, automated AI segmentations could enable more accurate lesion localization and streamline workflows through consistent contouring and compatibility with fusion biopsy platforms,” the researchers wrote in the abstract.
Catch this session to get more information.
