The Computer-Aided System for CArdiovascular Disease Evaluation (CASCADE) has been developed for streamlined, automated analysis of carotid artery magnetic resonance imaging to measure atherosclerotic plaque burden and composition in vivo. The purpose of this investigation was to assess the performance of CASCADE compared with manual outlining.
Magnetic resonance images were obtained from 26 subjects with 16% to 79% carotid artery stenosis by duplex ultrasound who were imaged twice in a 2-week period with a multiple-slice, multiple-contrast magnetic resonance imaging protocol as part of the Outcome of Rosuvastatin treatment on carotid artery atheroma: a magnetic resonance Imaging ObservatioN trial. Manual outlining was used to identify the boundaries of the lumen, wall, necrotic core (NC), and calcifications. After 6 months, the analysis was repeated using CASCADE. For each data set, the contours were used to compute the maximal normalized wall index (NWI; wall area divided by total vessel area), maximal wall thickness (WT), and the average NC and calcified (CA) areas per slice. Agreement between manual and automated reviews and the scan-scan measurement reproducibilities were evaluated.
Pearson correlation between manual and automated analyses was 0.94 for maximal NWI, 0.86 for maximal WT, 0.84 for NC, and 0.96 for CA. Intraclass correlation coefficients for manual and automated analyses were 0.90 and 0.97 for maximal NWI, 0.89 and 0.95 for maximal WT, 0.95 and 0.87 for NC, and 0.96 and 0.94 for CA, respectively.
Automated analysis tools are capable of providing accurate and reproducible measurements of carotid atherosclerotic burden and composition when compared with manually outlined results.
From the *Department of Radiology, University of Washington, Seattle, WA; †Department of Clinical Radiology, University of Munich, Grosshadern Campus, Munich, Germany; ‡Department of Cardiology, Juntendo University School of Medicine, Tokyo, Japan; and §Department of Surgery, University of Washington; and ∥VA Puget Sound Health Care System, Seattle, WA.
This study was supported by VPDiagnostics, Inc. (NIH SBIR grant no. R44-HL070576, AstraZeneca, LP and NIH training grant T32-HLO7838).
Reprints: William Kerwin, PhD, Department of Radiology, University of Washington, Box 358050, 815 Mercer Street, Seattle, WA 98109 (e-mail: email@example.com).