This chapter introduces a spatiotemporal statistical shape model (stSSM) using brain MR image which will represent not only the statistical variability of shape but also a temporal change of the statistical variance with time. The proposed method applies expectation-maximization (EM)-based weighted principal component analysis (WPCA) using a temporal weight function, where E-step estimates Eigenvalues of every data using temporal Eigenvectors, and M-step updates Eigenvectors to maximize the variance. The method constructs stSSM whose Eigenvectors change with time. By assigning a predefined weight parameter for each subject according to subjects’ age, it calculates the weighted variance for time-specific stSSM. To validate the method, this study employed 105 adult subjects (age: 30–84 years old with mean ± SD = 60.61 ± 16.97) from OASIS database. stSSM constructed for time point 40–80 with a step of 2. The proposed method allows the characterization of typical deformation patterns and subject-specific shape changes in repeated time-series observations of several subjects where the modeling performance was observed by optimizing variance.
Part of the book: Non-Invasive Diagnostic Methods