Comparison of digital boundary detection and semi‐automated analysis of left ventricular cine angiograms

A technique has been developed to analyze automatically 35 mm left ventricular cine angiograms with an image dissector and digital computer. The analysis procedure detected the ventricular boundary and aortic valve position, then calculated the projected left ventricular area and estimated volume, and performed a polar wall motion analysis as a function of time. The automated method was compared to a semi-automated method using a technician-operated electronic cursor linked to a mini-computer. The two methods were used to analyze a set of idealized ventricular models as well as normal and abnormal angiographic data. The volumetric and wall motion data for the idealized models calculated by the two methods were equivalent (r greater than 0.99). Volumetric analysis of the patients' ventriculograms produced correlation coefficients of 0.88-0.99. Corresponding wall motion data produced correlation coefficients of 0.72-0.94. The time necessary to analyze 30 frames of cine ventriculogram was approximately 30 minutes for each method.