Active shape models

Capturing the variability of fruit is useful in many ways, such as object detection and shape prediction. A statistical shape model is useful to predict the final size of a fruit, search for a correlation between stages of growth, or to estimate the volume of the fruit. Fruit package designers may use the shape model to evaluate the effect of the fruit on airflow characteristics and thus cooling uniformity.

This toolkit contains a collection of algorithms, implemented in C++ and Python, used for statistical shape modeling and analysis of 3D surface populations:

• 3D surface parameterization algorithms for disc, sphere, and cylinder topology
• 3D surface alignment
• Parameterization-based MDL correspondence optimization for disc and cylinder topology
• elasticity and model regularized 3D surface registration
• GUI for surface annotation
• Statistical hypothesis testing for surface-based shape analysis

Statistical Shape Models capture the shape variation of a training set of shapes and can be registered to an image of an object of the class they represent by simple adjustment of their parameters. We have applied statistical shape models of healthy tracheas to the assessment and stenting of tracheal stenosis. The idea is that a model with healthy tracheas only will not be influenced by local geometric variations typical of stenosis. The model produces a shape that is an estimation of the shape of the patient's trachea if it were not narrowed.

The airways comprise the hierarchical tubular structure that leads the air into the lungs, namely the trachea and the bronchi. A well know challenge in the pulmonology field is the automatic segmentation of the airways from tomographic images. We have developed a region growing algorithm in which the segmentation is iteratively bounded by cylinders. These cylinders limit the expansion of leaks, a common problem with region growing, and make them more easily detectable through the use of anatomical information about the airways.