[Insight-users] Advice on 2D image to 3D Mesh registration

[Carl Trapani]

Hi All,

I'm trying to understand if the approach I'm taking is feasible and 
would love any feedback from those more experienced and knowledgeable 
than I. I don't have much time to explore the numerous alternatives that 
exist and could really use some expert advice.

We are trying to build a 3D visualization system for gene expression 
spot data, much like the Allen Brain Atlas Brain Explorer. The idea is 
to display spots from cross sectional specimens within a 3D mesh model 
of the mouse cerebellum. Additionally, we want to show the spot data at 
different stages of cerebellar development. The essential problem is how 
to establish a spatial reference system that is consistent across 
developmental stages and then how to register the cross sectioned spot 
data to fit into that system.

Our approach is to develop 3D mesh models at each stage of development. 
These models will then act as fixed images to register the spot data 
sections to. Because the 3D mesh models only indicate the cerebellar 
boundary surface, while the gene expression image sections show spots 
for a specific genes within the entire cerebellar region, we cannot 
register the spot data directly against the model (right?). Our approach 
is to register 2D boundary images or masks, one sliced out of the model 
corresponding to the slice plane that created the spot data cross 
section, the other a manually traced spline around the spot data 
cerebellar structure. The resulting transformation from the 2D to 2D 
mask registration can then be applied to the spot data. Now, we hope to 
be able to stack these spot data sections, import as volume data and 
only have to re-scale and align bounding boxes to be able to fit the 
spots inside the model.

We are considering a landmark registration process for the 2D to 2D mask 
registration.

Thanks much for any of your thoughts - including the "it'll never work" 
ones. Also, pointers to papers or examples would be appreciated or 
references to materials/links (I have both Insight into Images and the 
Visualization Toolkit books - so reference that content if needed :-)

Thanks,
Carl Trapani