Influence of Data Parsing on Contrast Enhanced Ultrasound Exams

Abstract

To explore the influence of data parsing (either selection of frames at set time intervals or by an experienced sonographer) of contrast-enhanced ultrasound (CEUS) exams on physician diagnoses and confidence levels. Forty consecutive CEUS exams consisting of 10 cases each of indeterminate liver lesions, indeterminate renal lesions, renal cell carcinoma postablation follow-up, and hepatocellular carcinoma postchemoembolization follow-up were selected for analysis. Exams were parsed into sets consisting of five images selected by the performing sonographer and sets containing systematically stored frames every 10, 30, and 60 seconds. Three blinded physicians then reviewed the cine loop and each set of images in randomized order and provided a diagnosis and confidence level. For all clinical applications investigated, no statistically significant differences in diagnostic performance measures or reader confidence were observed between review of the entire cine loop and images selected by the performing sonographer (p > 0.42). Diagnostic performance at 10-second intervals did not show statically significant changes compared to the full cine loop review for all applications (p > 0.18), although reader confidence decreased. At 30-60-second intervals, both diagnostic performance and reader confidence showed statistically significant reduction compared to review of the full cine loop (p < 0.045). Transfer and review of large cine loops from CEUS exams represent a potential barrier to adoption within the United States workflows. This study demonstrates that images selected by a performing trained sonographer may provide the same value without the review time and data storage costs needed for full cine loop review. Parsing by time points reduced reader confidence and diagnostic performance.