Amygdala shape analysis and parametric surface visualization using iterative closest point algorithm and spherical mapping

摘要

The analysis of shape variability of anatomical structures is vital in a number of clinical applications, as abnormalities in shape can be related to the pathogenesis of certain diseases. This study is conducted for the amygdale shape analysis. The goal is three-fold: (1) develop a new framework of internal structure and surface parameter extraction and comparison for 3D shape data; (2) detect amygdale abnormalities in panic disorder using this technique; (3) freely apply various statistical methods and visualization of their results. To align the amygdala shape, Iterative Closest Point (ICP) algorithm is employed. Additionally, a fine-scale spherical mapping is used to generate various 3D shape parameters, including the radius from center to surface, Gaussian and mean curvature, and normal vector of the surface. Various statistical methods such as correlation, t-test, ANOVA, and ANCOVA with clinical parameters can be applied to compare the local shape differences between normal panic disorder patients and healthy comparison subjects. To apply this technique to clinical data, panic disorder patients and healthy volunteers were recruited and scanned by 1.5T MRI. Every amygdala was manually segmented on T1 MRI images and produced into the 3D surface model by Marching Cubes algorithm. The results indicate that group shape differences clearly exist in amygdala between healthy controls and panic disorder patients, which conform to clinical knowledge.