Project title: "Deformable Template Models and Shape Feature Descriptors for Brain Shape Morphometry" (DTMforBrainAnalysis)

Collaborating institutions: Computer Graphics and Visualisation Research Lab at the Korea Advanced Institute for Science and Technology (KAIST, South Korea)

About the project

The analysis of the shape morphometry of the brain structures is paramount in increasing our understanding on how the brain ages and changes with disease, and which factors may influence it. These, in turn, will ensure faster disease recognition and early diagnosis of age-related diseases, and better prognosis of other diseases that can dramatically affect the shape of brain structures like stroke. This project is dedicated to developing computational algorithms to accurately represent the shape of brain structures on individual basis and yet maintain point-to-point correspondence between the shape of the brain structures amongst individuals to allow both: individual and large-scale analyses.

The approach consists in developing deformable surface models using a “progressive surface deformation” of templates of brain structures to reconstruct the target shapes robustly while preserving the geometric details of initial models (template models) against size and shape variations across subjects. In addition, we also propose structural feature descriptors and anatomical landmarks, anatomically and geometrically defined on the template models, to investigate the shape characteristics and changes of the brain structures against clinical indicators. Proposed feature descriptors and landmarks guarantee comparability and consistency across subjects.

To date, the proposed methods have been validated hippocampus and brain ventricles of human data, collected from the Lothian Birth Cohort 1936 (LBC 1936) study, at The University of Edinburgh.

Associated publications

Please, view the list of publications on the DTMforBrainAnalysis project website

DTMforBrainAnalysis publications (external link)

Outputs

IMPORTANT NOTE: Developed by us algorithms are publically available, free of charge. You are free to use developed by us tools for research purposes. Hovewer, please include adequate citations and acknowledgments whenever you present or publish results that are based on it.

Group-wise Shape Analysis Tool

A software, publically available for download from the project website, which includes:

AverageImageConstruction
takes binary images of target structure, and outputs group-wise average image.
IndividualSurfaceReconstruction
takes template surface model & individual binary mask, and computes individualized surface model using progressive Laplacian surface deformation.
SurfaceAlignment
takes a template surface model & individualized surface models, and outputs an average surface model.
LocalDeformity
takes the average surface model & individualized surface models, and outputs a matrix (TXT format) containing the deformity values (the signed Euclidean norm of the displacement vectors) at each vertex.