In this paper, we present a method for predictive modeling of anatomic structures using canonical correlation analysis (CCA). Using this technique, certain anatomical structures, such as tumor-distorted structures, can be estimated from others by exploring the correlation between them, which has been determined from a set of training samples. Cortical surfaces and corpus callosum boundaries have been used to demonstrate the performance of the proposed method in predictive modeling. Applications of this method are in estimating brain tissues obscured by tumors and surrounding edema, in detecting abnormal structures, and in formulating alternate forms of statistically-based interpolation and regularization.