More CT scan import data

[mrxd]

Another 64-slice CT scan imported into Osirix and LW modeler and whittled down to 0.5% of original polygon data (currently about 200,000 polys).

With modeler you can select different structures and texture them (carotid arteries and jugular veins). It is quick and dirty, with little hope of good rigging or animation but anatomically accurate.

[JeffRutan]

This must have been a nightmare to sift through with all those stray points and so much data at once probably gave your computer a run for the money too!
Congrats on getting this far! How long did you spend on it so far?
The real challenge would be to write a plug in to do this kind of clean up automatically. ;-)

[mrxd]

The stray points only took a few minutes. The longest part of the whole procedure was using QEMLoss3 to decrease the polys to 5%. That took about 5-10 minutes of crunching. It was pretty quick and easy but that is because I was using a Quad G5. My older dual 1GHz G4 would chew on it for a long time.

[DrTWT]

very cool,

i've toyed with the idea of doing this but it seems to be a big project.

i'd love to see more.

[loki74]

If only there was a way to get it into a reasonable, animatable/riggable mesh...

anyhoo this is some pretty interesting stuff.

[mrxd]

The hard part is coming up with the DICOM data from the CT scan. Hopefully, the scan was done on a newer scanner and the study was done on a desirable body part.

Once you have the scan on a CD, use Osirix (freeware Mac medical image viewer) to make a shaded surface display model (you can vary the detail from a manageable 10% to a anchor-dropping 100%). You then export the model as an .obj.

Open the .obj you saved from Osirix with Lightwave Modeler and then use QEMLoss3 to reduce the polys even more (to 5%). You can easily select stray polygons and points. Then try highlighting and texturing whatever structures you want.

The model is very accurate (especially if you don't reduce the detail as much) but it is inefficient and messy, with little hope for useful animation.

Care must be taken not to reproduce a likeness of soemone who has not given their consent. Names and faces are HIPAA violations. I try to warp mine with the magnet tool.

[paulk]

. . . It is quick and dirty, with little hope of good rigging or animation but anatomically accurate.
Maybe I'm not seeing things correctly, but the attached image shows "extensions" on both sides of the upper and lower jaws. I assume they're "artifacts" from the conversion process. Or was the original subject trying to catch a bullet in his/her teeth at the time?

[mrxd]

those extensions coming from the mouth are streak artifact from the scan itself. The patient's dental work produces those spikey extensions and it is an annoying artifact on all CT scans with dense metal (dental work, hip replacements, dense barium, etc).

You won't have those if there is no metal in the scan.

[DrTWT]

For a while I played around with a Voxar workstation at the hospital (see http://www.barco.com/medical/en/downloads/3dimages.asp for a gallery of Voxar images.)

We never went to far for this since we couldn't get the images clean enough; we were mainly looking to use it for still illustrations and the like.

It would be cool if you could import CT data into LW in a usable form...

Here are a few images straight from Voxar

[Elmar Moelzer]

You guys might want to wait a few days and get to see some cool stuff for LW thats coming up ;)
CU
Elmar

[Elmar Moelzer]

well maybe I should give away a little teaser... just a tiny little one... cant hurt, right?
http://www.mediastudio-graz.com/volumedic/images/vm_torax_shadows01.jpg

[DrTWT]

You guys might want to wait a few days and get to see some cool stuff for LW thats coming up ;)
CU
Elmar

Elmar:

I did a little searching on the web for VoluMedic; this very well may be a must-have for me. Can you let out any additional information??? Feel free to PM me....

Please keep me up to date on this,

Thanks!

[Elmar Moelzer]

I did not really want to hijack mrxds thread here, but I it was too hard for me to keep my mouth shut ;)
Sorry mrxd.
I will let you all know when there is more to say than I can say now.
Be prepared for more info to get out in April though...
CU
Elmar

[DrTWT]

mrxd:

to get back on topic, any more renders to post? also, would love to see any wires you have; both pre- and post- poly reduction.

[mrxd]

here are a couple of meshes of this render...

[mrxd]

Here are a couple of older ones I did. The face is semi-transparent (nose & cheeks distorted a little to prevent recognition) and the same person's actual skull is rendered inside.

The foot was rendered with clear, transparent skin. The plaster splint (green) and bedsheet (blue) are from the actual scan data too. I did not make any part of the model except to reduce polygons in Osirix and LW Modeler.

This kind of modeling would probably be perfect for medical illustration but not so hot for animation (see mesh).

[mrxd]

Elmar,

That definitely looks like something I would like to fiddle with. I only hope that it will be a Mac compatible plugin.

[Elmar Moelzer]

Hey!
No Mac- version right now, sorry (we have one planned though, but that might take a while, especially since there is still problems with OpenGL2.0 support on Mac from what I know).
To get back on topic: I am a little curious about the weird look of the spot where the clavicula meets the manubrium. Was that part simply missing in the dataset or does the patient have an injury there?
CU
Elmar

[mrxd]

That defect in the medial clavicles is the same reason there are holes in the front of the maxillary sinuses. The bone is very thin (cancellous bone in the clavicle heads) there and the density was not high enough to be selected to make the shaded surface model. I could get rid of it by lowering the threshold for the density to make the model in Osirix but then I would start to collect a lot of other undesireable bits and pieces that I would have to spend more time getting rid of later.

If I spent more time on it I could make these artifacts minimal but these models were made with a minimum of effort on my part (just a lot of computing effort).

Cheers

[DrTWT]

This kind of modeling would probably be perfect for medical illustration but not so hot for animation (see mesh).

That foot and ankle actually look quite good. We do a lot of animations of procedures in which the patient is stationary and the equipment moves, so this might be quite useful.

Do you know of something similiar to Osirix that works in Windows?

[Elmar Moelzer]

Ahh ok, did not know how Osirix handles the mesh- generation.
In VoluMedic we use LWs Gradients and some other functionality for defining the surface of the mesh. So far this has proven to work quite well.
We still want to combine this with some other methods like region growing in later versions to make segmentation even better.
We also have another method ready that wont be in version 1.0 though (some non technical issues are preventing us from releasing this right now).
The foot looks indeed pretty good though!
CU
Elmar

[mrxd]

I don't know of any specific programs for PCs that are like Osirix but I am sure there are several. I would "Google" it and see. I seem to recall several PC DICOM viewers but I don't know if they create and export models.

You might want to try Versiontracker.com and search under 'DICOM' or 'medical image viewer'.