Can the same edge-detection algorithm be applied to on-line and off-line analysis systems? Validation of a new cinefilm-based geometric coronary measurement software

In the Cardiovascular Measurement System (CMS) the edge-detection algorithm, which was primarily designed for the Philips digital cardiac imaging system (DCI), is applied to cinefilms. Comparative validation of CMS and DCI was performed in vitro and in vivo with intracoronary insertion of stenosis phantoms in anesthetized pigs. The "obstruction diameter" (OD) was measured at the artificial stenoses visualized by angiography with calibration at the isocenter (ISO) and catheter calibration (CATH) and compared with the true phantom diameters. A clinical comparison of OD, reference diameter (RD), and percentage diameter stenosis (DS) was performed on 70 corresponding images from post-PTCA angiograms. In vitro, OD (CMS) yielded an accuracy of 0.18 +/- 0.14 mm with 100% (correlation coefficient: r = 0.97, y = 0.06 + 0.75x, standard error of estimate [SEE] = 0.09) and 0.19 +/- 0.15 mm with 50% contrast (r = 0.94, y = 0.02 + 0.81 x). OD (DCI) yielded an accuracy of 0.11 +/- 0.06 mm with 100% (r = 0.99, y = -0.03 + 0.91 x, SEE = 0.05) and 0.24 +/- 0.13 mm with 50% contrast (r = 0.94, y = 0.29 + 6.69 x, SEE = 0.12). In vivo, OD (CMS) yielded an accuracy of 0.18 +/- 0.23 mm with ISO (r = 0.89, y = 0.02 + 0.83 x, SEE = 0.22) and 0.26 +/- 0.24 mm with CATH (r = 0.89, y = 0.06 + 0.72 x, SEE = 0.19). OD (DCI) yielded an accuracy of 0.08 +/- 0.15 mm with ISO (r = 0.96, y = 0.08 + 0.86 x, SEE = 0.14) and 0.18 +/- 0.21 mm with CATH (r = 0.92, y = 0.09 + 0.76 x, SEE = 0.17). The clinical comparison showed reasonable agreement for OD only (r = 0.81, y = 0.26 + 0.81 x, SEE = 0.29). Transformation of an edge-detection algorithm from a digital to a cinefilm-based system can lead to impairment of measurement reliability.