Orthogonal shape modes describing clinical indices of remodeling

Abstract

Quantification of the left ventricle (LV) shape changes (remodeling) is of great importance for therapeutic management of myocardial infarction. Orthogonal shape modes derived from principal component analysis (PCA) often do not describe clinical remodeling indices. We developed a method for deriving orthogonal shape modes directly from any set of clinical indices. Cardiac magnetic resonance images of 1,991 asymptomatic volunteers from the MESA study (age 44−84, mean age 62, 52% women) and 300 patients with myocardial infarction from the DETERMINE study (age 31−86, mean age 63, 20% women) were obtained from the Cardiac Atlas Project. Clinical indices of LV size, sphericity, wall thickness and apical conicity were calculated. For each index, cases outside two standard deviations of the mean, but within one standard deviation for all other indices, were chosen as a representative subgroup. Orthogonal modes were defined sequentially, using the first principal component of each subgroup. At each step, the contribution of the previous mode was removed mathematically from the shape description, similar to Gram–Schmidt orthogonalization. Correlation analysis and logistic regression were performed to show the effectiveness of these features to characterize remodeling due to myocardial infarction.