3D reconstruction using section images

Hi everyone, I’m quite new to Panda3D and I have just started to work for for a medical company. They need me to implement an application that needs to be able to recontruct a jaw in 3D from the DICOM image files of the jaw’s sections. So, I’m wondering if that kind of application can be done with panda3D/python, and I would need to be in real time so we can move the camera and interact with the 3D model. What do you think, would it be possible?


Could you post a few section pics so we can get an some idea on how to do it?

It should be quite posible techicaly, though you are going to need some skills to pull it off.

Or just pay me to do it. :slight_smile: (Just kidding!)


Panda can build a terrain mesh from image files, you may be able to use that feature to do what you want.
See the manual pages starting at:

If I get it right, we are talking about a series of cross-sections.

The easy way would be to make a series of planes, and but every image on it. The only hard part would be adding alpha to the images.

Here’s a mock-up(I’ve used the images as alpha and colour so it’s ghost-like):

and the model + textures:

But this is just a visualization, It’s not a real 3d model. I think making a model from a series of 2d images is not something out-of-the-box for Panda3D.

It would not be very hard to use PIL to create alpha maps based open the data.

thx people, what I need is a 3D model, not a bunch of alpha planes, so I can’t do it that way. But thanks for all the help and putting the images and stuff. I’ll try to look up on the terrain generation manuals to find some clue if I can really take advantage from that. :slight_smile:

Its is posible to do what your talking about, its just very, very hard.

If you could post some images I will see what I can do.

the problem is , calculating the 3d model from the input images. i suggest you search for some open source medical software to do that. once you have a 3d model, panda can display it.

Here is how I am thinking of doing it:

Use PIL to create a set of vertexes and lines around the part of the image that is not empty.

Take this and extrude it to the correct height. (How far each image is spaced apart.)

After doing this for each image, the models are stacked on top of each other.

@Hiranga, Does the model need to be created in real time from images, or do you only need to interact with it in real time?

You should take a look at VTK if you haven’t already: vtk.org/

I have no experience with it but it has Python bindings and I believe it can read DICOM files. The right tool for the right job. :wink:

edit: vtk.org/Wiki/VTK_FAQ#How_can … files.3F

Some folk recomended me to use DeVIDE, a framework that seems it should help me with all that technical stuff. As Hollower said, it uses VTK/ITK libraries and implements a kind of friendly user interface that lets you “program” adding blocks to a canvas and linking them with each other as if you were calling functions and passing parameters. In addition it lets you add your own pieces of code in order to create more specific modules. So by now I will try this framework and see if it really can be used to run this project. Thank you all.

its quiet easy.

take a look into pxLJunks revenge III,which comes quiete nice along.

there you will find for a very low amount of money (~40 euro) exactly what you are looking for.

[url]a easter present]

if you watch there, for a example the test terrain (and other models) i done. its used the same procedere, fully shaded with specular and ready to use your virtual dental drill for drilling holes in. :smiley:

I think steaveford is a robot. I looked at his other posts and they are similar nonsense appearing to be constructed from pieces of what others wrote. They usually insert spam links but I guess maybe this one is broken :unamused:

You’re right. Probably one of those bots that makes posts that seem to make sense, and after a while, they edit their signature to include a spam link.
Thanks, user and posts deleted.