Image display settings (window and level) have a substantial impact on measurements of coronary artery and plaque dimensions in computed tomography (CT), and their influence on measurement accuracy has not been systematically evaluated. We analyzed the influence of window width/level settings on the accuracy for determining cross-sectional lumen and outer vessel diameters in contrast-enhanced CT angiography compared with intravascular ultrasound (IVUS).
We evaluated the data sets of 35 patients. Coronary CT angiography was performed as part of a research protocol before invasive coronary angiography. A contrast-enhanced volume data set was acquired using a dual-source CT (DSCT) scanner (Siemens Healthcare, Forchheim, Germany). Intravascular ultrasound was performed using a 40-MHz IVUS catheter (Atlantis, Boston Scientific Corporation, Natick, Mass) and motorized pullback at 0.5 mm/s. One hundred exactly corresponding sites within the coronary artery system were identified in both DSCT and IVUS using bifurcation points as fiducial markers. In DSCT data sets, multiplanar reconstructions (0.75-mm slice thickness) were rendered orthogonally to the centerline of the coronary artery at each of the 100 sites. Computed tomographic images were displayed using 4 previously published settings (700/200, 700/140, and 500/150 Hounsfield units [HU], and 1 HU/65% of the mean luminal intensity [HU] and 155%/65% of the mean luminal intensity [HU] for window width/level) as well as with a visually adjusted setting for subjectively optimal lumen and outer vessel area measurement. Coronary lumen and cross-sectional vessel areas were manually traced using all 5 display settings and compared with IVUS measurements.
Concerning cross-sectional vessel area measurements, correlation was close and significant compared with IVUS using all settings (r ≥ 0.93, P = 0.01 for all settings). Bland-Altman analysis revealed a good agreement between both modalities with a systematic bias toward overestimation in CT. Least bias was demonstrated using the setting 155%/65% of the mean luminal intensity for window width/level, with a mean (SD) difference of 0.2 (1.73) mm2. For the measurement of the luminal area, the window setting using a width of 1 HU and a level of 65% of the mean luminal intensity showed the lowest correlation to IVUS (r = 0.85), with a systematic bias toward underestimation of the lumen in CT. Bland-Altman analysis revealed a moderate agreement with a mean (SD) difference of −2.1 (1.6) mm2. For all other settings, a very close correlation was observed (r > 0.9, P = 0.01), and Bland-Altman analysis revealed a slight trend toward lumen underestimation in CT, yet with a good agreement. The least bias was demonstrated using the setting 700/200 HU for window width/level with a mean (SD) difference of −0.1 (0.9) mm2.
Previously published window settings and visually adjusted window setting correlate very well with IVUS measurements regarding coronary artery cross-sectional and luminal area, with a better agreement for luminal area measurements. A systematic bias toward overestimation of vessel area in CT was observed as well as a slight trend toward lumen underestimation. This bias was least for vessel area measurement using 155%/65% of the mean luminal intensity (HU) for window width/level, whereas for luminal area measurement, the setting 700/200 HU for window width/level yielded the least bias.
From the Department of Cardiology, University of Erlangen, Erlangen, Germany.
Received for publication March 12, 2010; accepted August 4, 2010.
Reprints: Mohamed Marwan, MSc, Department of Cardiology, University of Erlangen, Ulmenweg 18, 91054 Erlangen, Germany (e-mail: email@example.com).
The study was supported by Bundesministerium für Bildung und Forschung (BMBF), Bonn, Germany (grant BMBF 01 EV 0708).