Abstract
Objectives: A critical review was undertaken to synthesise current research on the diagnostic accuracy of artificial intelligence (AI) compared with radiologists in identifying and classifying pulmonary nodules in computed tomography (CT). Secondary aims were to evaluate scan reading time and assess study design of included articles.
Methods: Literature was searched using Medline, PubMed and CINHAL databases and screened according to inclusion and exclusion criteria. Data of interest was extracted from selected studies and study quality was assessed using CASP for diagnostic test studies.
Results:Four studies with 3,746 pulmonary nodules from 40,586 participants were included in this critical review. Studies were deemed of similar standard. AI intervention showed greater identification sensitivity than radiologists (90.1% vs 64.8%, P<0.001), however, the false positive rate per scan of AI was considerably higher than that of radiologists (1.72 vs 0.58). Regarding nodule classification, AI and radiologists yielded highest...
Sign in to your account
Log in
This content is available to registered users only. Please sign in to continue.
If you are not yet registered, please create your account on www.sor.org and verify your membership status. Once your account has been set up, return here and sign in to access the content.
Set up your account