Fully automatic detection and measurement of abdominal aortic aneurysm using artificial intelligence. (14th October 2021)

• Record Type: Journal Article
• Authors: Fujiwara, J F; Orii, O M; Araki, K A; Ogura, M O; Ito, T I; Oyamada, K O; Morino, Y M; Yoshioka, K Y
• Is Part Of: European heart journal. Volume 42(2021)Supplement 1
• Publication Date: 2021-10-14
• Subjects: Artificial Intelligence (Machine Learning, Deep Learning); Cardiology -- Periodicals; Heart -- Diseases -- Periodicals; 616.12005
• DOI: 10.1093/eurheartj/ehab724.3070
• Languages: English
• ISSNs: 0195-668X

Abstract

Abstract: Background: Abdominal aortic aneurysm (AAA) is a life-threatening disease, and the decision to treat relies on the exact detection and size of AAA, which is time-consuming and prone to high inter-reader variability. Artificial intelligence (AI) has revealed new insights into the management of AAA, but it needs large amounts of contrast-enhanced CT data. Moreover, there has been few studies about the fully automatic detection and measurement of AAA. Purpose: The aim of this study is to evaluate the accuracy of AI for the detection and measurement of AAA using small amounts of non-contrast CT. Methods: We retrospectively collected 145 non-contrast CT scans of suspected AAA. The first step was to manually segment the 30 non-contrast CT scans to create training data for the AI, and the second step was to improve accuracy using 9 of non-contrast CT scans. Image processing is used to identify the AAA area and automatically measure the size. The AAA region was identified from the amount of change in the minor-axis, and the maximum minor-axis diameter was calculated using elliptical fitting. To evaluate the diagnostic utility of AI for the detection of AAA, the sensitivity and positive predictive value were calculated referred for the diagnostic report. The reproducibility in the size of AAA was assessed using intraclass correlation coefficients (ICCs) between the diagnostic report and AI. Results: Among 145 of non-contrast CT scans, 111 had AAA. The sensitivity and positive predictive value of AI for the detection of AAA were 94.6% and 53.8%, respectively (Picture 1). The size of AAA calculated by the AI (42.5±8.8 mm) showed a strong correlation with the those of diagnostic reports (44.6±8.4 mm; ICCs = 0.97) (Picture 2). Conclusions: AI represents a useful tool in the fully automatic detection and measurement of AAA using small amounts of non-contrast CT. FUNDunding Acknowledgement: Type of funding sources: None.