Utilization of Artificial Intelligence–based Intracranial Hemorrhage Detection on Emergent Noncontrast CT Images in Clinical Workflow

• DOI: 10.1148/ryai.210168
• Publication Date: 2022-02-09T14:55:27Z
• Authors: Muhannad Seyam; Thomas Weikert; Alexander Sauter; Alex Brehm; Marios-Nikos Psyc...; Kristine A. Blackham
• Keywords: Turnaround time

Authors implemented an artificial intelligence (AI)-based detection tool for intracranial hemorrhage (ICH) on noncontrast CT images into emergent workflow, evaluated its diagnostic performance, and assessed clinical workflow metrics compared with pre-AI implementation. The finalized radiology report constituted the ground truth analysis, examinations (n = 4450) before after implementation were retrieved using various keywords ICH. Diagnostic performance was assessed, mean values their respective 95% CIs reported to compare (report turnaround time, communication time of a finding, consultation another specialty, in emergency department). Although practicable observed overall ICH 93.0% accuracy, 87.2% sensitivity, 97.8% negative predictive value, yielded lower rates specific subtypes (eg, 69.2% [74 107] subdural 77.4% [24 31] acute subarachnoid hemorrhage). Common false-positive findings included postoperative postischemic defects (23.6%, 37 157), artifacts (19.7%, 31 tumors (15.3%, 24 157). such as communicating critical finding (70 minutes [95% CI: 54, 85] vs 63 55, 71]) average reduced implementation, future efforts are necessary streamline all along chain. It is crucial define clear framework recognize limitations AI tools only reliable environment which they deployed. Keywords: CT, CNS, Stroke, Diagnosis, Classification, Application Domain © RSNA, 2022.