Tuesday, December 3 | 1:40 p.m. - 1:50 p.m. | T6-SSCH06-2 | Room E451A
Research conducted in the U.K. will be presented in this session that evaluated the performance of an AI algorithm designed to autonomously report high-confidence normal chest x-ray studies. With the current high burden of chest x-ray reporting on radiologists, such novel methods could be useful to streamline radiological assessments.
Presenter Sonam Vadera, a medical student at University Hospital Birmingham, and colleagues included 82,761 chest x-ray clinical reports across various outpatient and inpatient sources and used a separately validated natural language processing (NLP) tool to classify them. The group then analyzed concordance between the classifications of ChestLink AI software (Oxipit, Lithuania) and the clinician reports to evaluate accuracy and reliability.
Among the x-rays reviewed, 37,048 were normal scans based on clinician opinion. ChestLink was able to confidently report 9,821 (26.5%) of these as normal, effectively removing them from further assessment and expediting the workflow, according to the study. Overall, discordance was observed in 426 (4.3%) cases, where the software classified the x-ray as normal while clinicians reported abnormalities. Among these discordant cases, 15 (0.15%) were deemed clinically significant.
The findings highlight the potential of AI tools in enhancing the efficiency of chest x-ray reporting, according to the group. Check out this scientific session on chest radiography to learn more.
