DL May Assist With Detection of Abdominal AD

Background: Acute abdominal aortic dissection (AD) is a serious condition, and early detection using ultrasound (US) can speed up detection and improve prognosis. Although diagnosis with CT is the gold standard, US is a valuable tool for evaluating acute abdominal AD due to its wide availability, bedside applicability, and ability to provide fast and real-time results. Objective: To evaluate the ability of deep learning (DL) to diagnose abdominal AD on US images, which may aid diagnosis of AD by novice radiologists or non-professionals. Methods: 374 US images from patients treated before June 30, 2022, were evaluated and classified as either AD-positive or AD-negative images. Of these images, 90% were used as the training set, and 10% were used as the test set. In this study, a Densenet-169 model and a VGG-16 model were used and compared with 2 human readers. Results: DL models demonstrated high sensitivity and area under the curve for diagnosing abdominal AD in US images, with overa more...