pouët.net

Volume rendering in radiology: some new videos

category: general [glöplog]
Thats great and all, but autumn's not a month.
Quote:

That looks really good Navis. Great rendering, and the application looks really usable.

I'm wondering something though: would it be possible to simulate an operation with this? i.e. mark an incision on the skin, then have the software discard the area around the incision to show the layers beneath? That could be pretty useful I suspect.


I personally don't believe that physical modelling of incisions and emulation of muscle/fat/bone behaviour under external forces is something that has much chance in the real world (i.e. applied surgery), for many reasons. But removal of structures (as in, removal of a set of interconnected voxels) is very much doable; although again it is questionable whether it has much clinical use.

This is my opinion, I'm not a surgeon although I've worked alot with them over the years. I think that they want to be able to do 5 things properly rather than struggle with 50 half-baked ideas.
added on the 2009-12-06 20:27:43 by Navis Navis
Navis: I personally think it would be nice to see some 'medical influences' in your next (big) demo.. go for it! :)
added on the 2009-12-06 21:07:05 by magic magic
magic: BB Image
Medical goatse FTW
added on the 2009-12-06 23:25:23 by doomdoom doomdoom
wow, awesome. i read an article about medical rendering in an earlier issue of digital production this year. this is the way to go :)

i also saw that at the institue of robotics and mechatronics (DLR) when i worked there, even with force feedback experiments for realtime surgery assistance. quite cool, but if it was sooooo experimental.

i'm sure it's fun to do this stuff. (not for me, i'm totally hypochondriac... i guess i woud bleed out :)

i happened to get in touch with this stuff after beeing hit by a car and several snowboard injuries, and was always like "yeah, that's cool rendering of my inside :)"
added on the 2009-12-06 23:42:45 by prost prost
BB Image Navis will cure the cancer!
added on the 2009-12-07 02:17:10 by aftu aftu
either that or he'll make it look all shiny :)
added on the 2009-12-07 03:13:00 by Gargaj Gargaj
it musnt have been very clear, but my post was referring to the very first post....
Also, sorry :)

The thing seems quite great though.
Mah Li: LOL :D!!
added on the 2009-12-07 06:27:26 by panic panic
Quote:
I'm wondering something though: would it be possible to simulate an operation with this? i.e. mark an incision on the skin, then have the software discard the area around the incision to show the layers beneath? That could be pretty useful I suspect.


Well, years ago I talked to some of my physician friends about it. They all told me that they need to feel the real touch of things to be something useful for practices. Here at Spain, they spend a lot of years for the studies and then some 5 years more for practices.

So, maybe it would be useful at the time simulations can be done with enough realism it is visual and touch/pressure identical to reality... but this is very far to be real... much more in realtime.

I'm not sure if this is a good example but... to drive a car, for example, you might learn something by using a simulator, but in the end, you need to drive a real car to fully learn, and for sure you will learn a lot of more when you drive the real car.

By other hand, the renderings. For learning medicine, they all told me that good illustrations are better. And that is easy to understand: in a hand-made illustration, you can detail something as much as you want, making as clear as you want the important details, but also simplify unimportant things. This is the same as for every technical illustration.

So, what can be the biggest utility for medical image rendering? Radiology... what it is used currently for.

Yesterday I've been taking to one of my doctor friends, she is in the radiology department, and I told about how cool is to be able to render CTs in 60 fps with the current GPUs. Then I told her that, in any case, I think that for medical purposes, it doesn't matter if you render it at 4-5 fps or 100 fps... so no matter how cool it could be, it looks as something not really useful in the medicine field. She answered in a very clever way, it was something like "today, these developments are not directly useful for medicine, but when something new appears in the field, in some years it maybe gets applied to real medicine and gets useful"
added on the 2009-12-07 08:24:09 by texel texel
texel:

Volume rendering and CAD are very much useful (and have been for years) in many areas of radiology. Take for example virtual colonoscopy: granted, doctors can do this without fancy "travelling through colon" graphics, but it will take them more time, effort and skill.
added on the 2009-12-07 10:52:57 by Navis Navis
Hey guys,
I´m a consultant in Trauma & Orthopaedics and I must say that in our field of medicine, especially for complex trauma of, let´s say, the pelvis or the head of the tibia, there is a great need for visualisation of the bony structure in 3D to properly "put things back together". The fact that one can also visualise the soft tissues in order to plan one´s operative approach and "dissect" preoperatively is a mighty tool. We have rudimentary 3d reconstruction in some of our PACS systems from GE/Philips/Siemens, but none of the freedom in visulisation I saw here. Usually we can only turn along 1 axis as the data is only available as a succession of jpgs, not a fully transformable 3D set.
BTW navis, you´ve got mail ;-)
Never thought it would go from my old c64 to this ;-)
added on the 2009-12-07 13:14:54 by mct mct
thks. Now a couple of questions to the genies of volume rendering (iq?):

I have done subsurface scattering in the 3D graphics pipeline using a blurred depth map difference thingie. I guess you can do the same in raymarching based rendering, but is there a faster/better approach?

Ambient occlusion: should I bother or not (considering that these CT dataset have an element of ambient occlusion inherited from the scanning process - or so I think)

added on the 2009-12-07 15:21:52 by Navis Navis
Navis: We're doing AO in our volume renderer at work. We're not doing any sort of subsurface scattering, because it would be completely useless for our customers (Oil & Gas).
added on the 2009-12-07 15:25:05 by kusma kusma
Navis: I advice you so much to use AO, and, even better, indirect lightning if possible. The results are much more visuallly pleasant and the 3d is much more clear with it

http://www.romancortes.com/blog/voxel-head-in-flash/

Also, it can be done realtime I guess, there is a paper about using some kind of ssao to enhance 3d depth perception for CT scans... try to search for it, it is very interesting
added on the 2009-12-07 15:27:22 by texel texel
Naivs/texel: I wasn't thinking so much about "simulated operations", more along the lines of "if I go in here, will I sever this artery or that nerve". Could be useful in planning an op perhaps.

Navis: I'd say definitely add AO or at at least SSAO, it really helps to show the depth of the image. Also, dunno if you saw this experiment I did a while back: http://www.youtube.com/watch?v=eD-PD1EOxlE How about importing some demos into your app? :)
added on the 2009-12-07 15:49:06 by psonice psonice
this is all good, but:

This assumes that you have a clear-cut surface with opacity=1.0 doesn't it? How would it behave with opacities that range over the full spectrum. Most importantly, am I right believing that there is an element of AO *already* in CT scans? I base this on the following assumption that X-rays that cross a locally heavily occluded voxel of intensity A (lets say water) will attenuate more than one with the same intensity A that isn't occluded. I wonder if that is the case or not.
added on the 2009-12-07 15:53:21 by Navis Navis
Navis: we're just counting everything with high-enough alpha as occluders. It still helps to visualize where the different surfaces are compared to each other, even though it's nowhere near physically correct (our thingie is a crude SSAO-ish hack, but it's pretty fast and most importantly works pretty well with volume data in our use-cases).
added on the 2009-12-07 16:00:23 by kusma kusma
Jum... the AO question is very interesting... but I have no idea of it. I assume it might be true for x-ray, but what about MRI?
added on the 2009-12-07 16:55:46 by texel texel
Imho, MRI resolution is too crude/noisy to be used in any mode other than MIP. And you probably don't want to mess too much with MRI anyway. The physical correspondence of intensity is way more complex than with CT (taking into account weighting factors, i.e. T1 T2 PD).

I have yet to come across an MRI dataset that offers *more* insight into the data when volume rendered rather than viewed in MPRd slices with proper windowing levels.

I'm also pretty sure than my explanation of the AO effect in CT (*if* it is valid) does not apply to MRI.

added on the 2009-12-07 17:11:00 by Navis Navis
Any AO in the original data isn't showing up the way I'd expect to see AO in the render. If you do even cheap SSAO in the renderer it'll help to show the depth much more clearly. How you'd do it with a volume + transparency I don't know.. like kusma said having a simple cutoff might work well enough.
added on the 2009-12-07 18:34:32 by psonice psonice
that is the kind of AO effect I was talking about in CT:

www.asd.gr/~navis/feet.jpg

The image has no lighting whatsoever, and has a transfer function of an S-curve. All intensities (including =0) have >0 (albeit very small) opacity value.
Tou can see the darkening around gaps (maybe the partial volume effect on CTs is another explanation of the effect. Or that my eyes see AO in everything these days).

added on the 2009-12-07 19:21:21 by Navis Navis
may we expect a shiny proud penis in the next ASD prod ? :-)
added on the 2009-12-07 21:09:56 by aftu aftu

login