Statistical size-and-shape analysis of elastic curves for brain tumors

초록

Shape analysis is an active research topic in various application areas. To analyze shape from objects of interest, rigid motion and size are often removed. However, the size can also be regarded as an important feature as well as shape to discern the object. We propose a statistical approach to analyze size-and-shape of elastic curves with application to brain tumors in medical imaging. We provide tools for computing and visualizing descriptive statistics and implementing principal component analysis, which can simultaneously handle size and shape of curves. The proposed framework is illustrated through a dataset from radiologic images of patients with glioblastoma multiforme (GBM) brain tumor. We additionally demonstrate cluster and survival analysis based on size-and-shape of the brain tumor associated with clinical and genetic variables.

키워드