Quantitative regional curvature analysis: a prospective evaluation of ventricular shape and wall motion measurements

To overcome the assumptions and approximations mandated by the use of traditional wall motion methodologies, a method was recently developed for measuring ventricular shape based on quantitative curvature analysis of ventricular outlines. This study was designed to assess prospectively the performance of this algorithm, to compare it to traditional wall motion measurements (centerline method), and to determine the comparative degree to which each method mimicked the interpretation of wall motion by clinical observers. Semiquantitative visual grading of regional function in 52 patients was performed by four independent observers on two occasions. Anterior, apical, or inferior segments were judged to be normal (0 points) or abnormal (1 point) based on viewing nonrealigned, end-diastolic and end-systolic ventricular silhouettes from cineventriculograms obtained in the 30-degree right anterior oblique projection. Each segment was assigned a collated score ranging from 0 (all observers felt the region was normal on both readings) to 8 (all observers felt the region was abnormal on both readings). Quantitative regional curvature analysis and wall motion analysis (centerline method) were performed. Quantitative shape and wall motion scores correlated equally well with the semiquantitative visual scores. When a visual score of greater than or equal to 4 was used to designate an abnormal segment, both quantitative approaches demonstrated comparable sensitivity, specificity, and concordance rates. Both methods achieved optimal performance when maximum and minimum deviations from normal were recorded. Under these circumstances, the shape analysis demonstrated a greater concordance with the clinical diagnosis than did wall motion analysis (99% vs-93%, p less than 0.04).(ABSTRACT TRUNCATED AT 250 WORDS)