KMID : 0385920240350020165
|
|
Journal of the Korean Society of Emergency Medicine 2024 Volume.35 No. 2 p.165 ~ p.174
|
|
The trial of application for interpretation on brain computed tomography by emergency medicine residents assisted artificial intelligence algorithm-based solution
|
|
Kim Dong-Eok
Seo Young-Woo Ko Seung-Hyun
|
|
Abstract
|
|
|
Objective: This study examined the efficacy of artificial intelligence (AI) algorithm-based diagnostic assistant solutions in the interpretation of brain computed tomography (CT) by emergency medicine (EM) residents.
Methods: This study included 350 patients who visited a local emergency medical center over 5 months and underwent brain CT scans. EM residents initially interpreted the patients¡¯ scans. A second interpretation was performed using an AI algorithm-based solution. The initial and second interpretations were compared with that of a radiology physician.
Results: The first interpretation by EM residents showed agreement in 318 cases (90.9%), while the second, assisted by an AI algorithm-based solution, showed agreement in 308 cases (88.0%). The first interpretation had an accuracy, sensitivity, and specificity of 93.1%, 43.9%, and 99.7%, respectively, and the second had an accuracy, sensitivity, and specificity of 92.0%, 39.0%, and 99.0%, respectively (P<0.001). Most of the discrepancies observed in the first and second interpretations were classified as Grade 1.
Conclusion: The interpretations assisted by the AI algorithm-based solution resulted in lower accuracy and higher discrepancy rates than independent interpretations by EM residents. The AI algorithm-based solution provided efficacy in accurate interpretation depending on the cases. Further study will be needed to address the weaknesses of the function and utility of AI.
|
|
KEYWORD
|
|
Computed tomography, Image interpretation, Emergencies, Artificial intelligence
|
|
FullTexts / Linksout information
|
|
|
|
Listed journal information
|
|
|
|