Join our MCAD Central community forums, the largest resource for MCAD (Mechanical Computer-Aided Design) professionals, including files, forums, jobs, articles, calendar, and more.
How was the point cloud created, what method are you using to import it, and how do you know it is all positive to begin with? Are you able to share the files so we can troubleshoot?
The point clouds are generated in a MATLAB code that I wrote, so I know that the co-ordinates are all positive. Just to double check I opened one of the cloud files in word pad and yep they are all positive.
I have found that the problem doesnt occur if I use 3D Sketch to trace the contours in the Point Cloud and only occurs if I create a reference plane and Sketch on the reference plane.
Now that we know the source is good, let's review your import procedure... Sounds like they are coming in as ascii text... are you importing it as a *.sldcrv file then?
Can you explain the "reference plane" statement? I don't understand how you are importing a 3D point cloudonto a 2D plane... Can you tell me more about your geometry? Even better, can you upload a picture of it?
Yep 2010. Do you have an 'Add-Ins' option in the 'Tools' pull down? If so, add 'ScanTo3D', this will give you the Point Cloud file options in the File OPen dialogue.
As you can see I have +x and +z, and a datum plane through the lowerCT slice iscoming in around +0.726in in the y... Is this what you are getting?
I've recently been playing around creating surfaces from MRI data, but I've never seen a slice turned into a point cloud like this before. Are you just taking measurements, or are you ultimately going to create 3D models from it?
Now I am beginning to understand what you're having trouble with. Your initial post was missing quite a bit of information!
The "-y" values you see on the left-hand side under "paramters" (it's actually "-x" for me) is just the local (x,y) coordinates on that particualr sketch plane, and the directions are dependant on how SW orients the datum plane when it is selected for sketch. I have found this to bequite arbitrary.
For your task, is the "-" sign really messing you up that much? I mean, it's just the values that are important. When you exit the sketch those local coordinates don't matter.
Also, what made you choose this method? How did you segment the image to create the point cloud?Have you investigated usinga medical image processing software, which can create the volume for you?
I foresee a lot of manual labor in trying to create a spline-outline for each slice, then creating surfaces from that... Just trying to help you acheive your overall goal with the least amount of labor/time.
The problem arises when I work with an otrthogonal plane say the YZ plane I get strange rotations around the Z axiz. Although the 3D Sketching is working out fine at the moment.
I choose the method mainly due to memory problems when working with high resolution CT images. This way I can down sample easily, cut and crop, threshold on the cheep with my own code. Also, some of the down sampling used in commercial software was not addequate for what I'm doing. I'm not interested in volume just the surfaces. The FE model uses shells and beams so no need for volume.
I also get alot more control over the surface generated in SW.
Yes, it's when you try to sketch on a datum plane. SW seems to arbitrarily re-orient the sketch plane, and I do not know of a workaround. I'd recommend you stick with the 3DSketch method.
Great answer. I've been using some open source doc image processer, and it can segment and create an stl for me. However, thesegmentation algorithm isn't doing what I want it to, and as a result the STL's are not very accurate, so I'm a bit stumped with it right now...
Can you share with me a little more about your CT-Matlab-point cloud process? (And what exactly are you simulating with shells/beams?) Email me at [email protected] or PM me, if you don't want it on the forum. Good Luck with your projects otherwise!
I'm modelling hearts valves. You can get segmentation programs that will smooth and denoise images to get better geometries, ScanIP springs to mind. But I have never been happy with them. Its not just the geometry that comes out clumsy but when I have used auo-meshing algorthms the meshes are just horrific.
CT > Matlab > Point cloud convertion is pretty straight forward. I convert the CT stack to Tifs then exctract the different planes. Matlab alows all sorts of resampling algorthms to be done or you can even write your own if you need certain areas to have particular quailities. To convert to point cloud I simply threshold the images to get rid of eth back ground and write an XYZ txt file with co-ords of the point over and under a certain value.
Thanks for the scanIP lead, I'll check that out. My major setback is that the source image is not that clear. No post-processing is going to fix that...
No auto-mesh program is going to give you the mesh you want, especially advanced goemetry like body organs. I'd put the onus on the automesher, not the geometry in that case.
Neat stuff with matlab, I've never gotten a chance to use it. I'd be curious to see your final surface as a result of this method. Feel free to share it when you're done...
Yep I'll post a pic when its done. What are you scanning with the MRI? MRI images are notoriously blurred unless you have a high power scanner with no movement.
Matlab's M-code is alot like C/C++. Have just modified my code to give me the contours only, so just shaved a few days of the reconstruction
I have an MRI volume of a shoulder joint, and I'm using it as a starting point for creating a CAD model. Unfortunately, I don't think the quality is good enough for me to do anything with it.
Congrats, I look forward to seeing the final model.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
