This paper proposes a new method for understanding the structure of populations of complex objects in the area of medical image analysis. The new methods require invention of approaches to the statistical analysis of a population of tree-structured objects. The approach is based on a metric in tree space. The metric provides a foundation for defining a notion of population center. In Functional Data Analysis, variation about the center is usually analyzed by Principal Component Analysis. Here an analog of PCA is developed for tree space. |