Paper: | SA-PM-SS1.7 |
Session: | The Use of Shape in Biomedical Imaging |
Time: | Saturday, April 8, 17:20 - 17:40 |
Presentation: |
Special Session Oral
|
Title: |
Characterization of Biological Growth Using Iterated Diffeomorphisms |
Authors: |
Ulf Grenander; Brown University | | |
| Anuj Srivastava; Florida State University | | |
| Sanjay Saini; Florida State University | | |
Abstract: |
Mathematical and statistical modeling of biological growth are important problems in medical diagnostics. We study a structured model, called Growth by Random Iterated Diffeomorphisms (GRID), that models growth by emphasizing its local nature. The cumulative growth is composed of several smaller deformations; each deformation models an active region by capturing deformation local to that region, and is characterized by a seed and a radial deformation pattern around the seed. The GRID variables -- seed locations and radial deformation patterns -- are estimated from observed images in two steps: (i) Estimate a cumulative deformation over an observation interval, and (ii) Estimate GRID variables using maximum-likelihood criterion from the estimated cumulative deformation. We demonstrate this framework using MRI image data of a rat's brain growth. |