A novel approach for quantitative analysis of intracoronary optical coherence tomography: High inter‐observer agreement with computer‐assisted contour detection

Objective: This study aims to examine observer‐related variability of quantitative optical coherence tomography (OCT) derived measurements from both in vitro and in vivo pullback data. Background: Intravascular OCT is a new imaging modality using infrared light and offering 10 times higher image resolution (15 μm) compared to intravascular ultrasound. The quantitative analysis of in vivo intracoronary OCT imaging is complicated by the presence of blood, motion artifacts and the large quantity of information that has to be processed. Methods: We developed a standardized, automated quantification process for intracoronary OCT pullback data with inter‐observer variability assessed both in vitro by using postmortem human coronary arteries and in vivo by studying simple and complex coronary pathology and outcomes following stent implantation. The consensus between measurements by two observers was analyzed using the intraclass and interclass correlation coefficient and the reliability coefficients. Bland–Altman plots were generated to assess the relationship between variability and absolute measurements. Results: In vitro OCT assessment was performed in nine postmortem coronary arteries. The time needed for semiautomated contour detection of a 15‐mm long coronary segment was ∼40 min. The absolute and relative difference between lumen area measurements derived from two observers was low [0.02 ± 0.10 mm2; (0.3 ± 0.5)% respectively] with excellent correlation confirmed by linear regression analysis (R2 = 0.99; P < 0.001). Similarly, in vivo measurements demonstrated a high correlation with the main source of inter‐observer variation occurring as a result of coronary dissection and motion artifact. The absolute and relative difference between measurements were 0.11 ± 0.33 mm2 (1.57 ± 0.05)% for lumen area (R2 = 0.98; P < 0.001), 0.17 ± 0.68 mm2 (1.44 ± 0.08)% for stent area (R2 = 0.94; P < 0.001), and 0.26 ± 0.72 mm2 (14.08 ± 0.37)% for neointimal area (R2 = 0.78; P < 0.001). Conclusions: Highly accurate computer‐assisted quantitative analysis ofintracoronary OCT pullbacks is feasible with low inter‐observer variability. The presented approach allows for observer independent analysis of detailed vessel structures, and may be a valuable tool for future longitudinal studies incorporating OCT. © 2008 Wiley‐Liss, Inc.