How to get most accurate 3D volume?

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

How to get most accurate 3D volume?

Luo, Jiajia
Hi all,
 
I have created several 3D models from Labelmaps on MRI scans. The muscle I am tracing is small and most accurate in the coronal view, however when labelmap is complete the model created is poor partly because of the 5mm gap between MRIs and partly because the lofting is incorrect. I therefore have to make 'fat' tracings in order to make the model look and loft correctly. I also have to use Laplacian smoothing function.
 
Questions:
1) If I trace accurately the model is full of holes and there is no correction function for this is there?
2) Which is the most accurate volume for the model (based on the 'fat' labelmap tracing) ie one at 5,10,20,30,40 Laplacian or Sinc? And what is the difference between Laplacian and Sinc (filter type) in 3D slicer?
3) In a single element model should I tick the joint smoothing box ( as this does make a differnce to the appearance ) or does this make it even less accurate?
 
Thanks,
 
Jiajia
 
 

_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject
Reply | Threaded
Open this post in threaded view
|

Re: How to get most accurate 3D volume?

swallace
I'm not sure what you are looking for in terms of improved accuracy, but
have a look at this newer algorithm, and give it a try:


http://wiki.na-mic.org/Wiki/index.php/RobustStatisticsSegmentation

It needs to be run in 3.5, and it does give nice results; it might help
remedy some of the smoothing issues you're are encountering.

Stuart



> Hi all,
>
> I have created several 3D models from Labelmaps on MRI scans. The muscle I
> am tracing is small and most accurate in the coronal view, however when
> labelmap is complete the model created is poor partly because of the 5mm
> gap
> between MRIs and partly because the lofting is incorrect. I therefore have
> to make 'fat' tracings in order to make the model look and loft correctly.
> I
> also have to use Laplacian smoothing function.
>
> Questions:
> 1) If I trace accurately the model is full of holes and there is no
> correction function for this is there?
> 2) Which is the most accurate volume for the model (based on the 'fat'
> labelmap tracing) ie one at 5,10,20,30,40 Laplacian or Sinc? And what is
> the
> difference between Laplacian and Sinc (filter type) in 3D slicer?
> 3) In a single element model should I tick the joint smoothing box ( as
> this
> does make a differnce to the appearance ) or does this make it even less
> accurate?
>
> Thanks,
>
> Jiajia
> _______________________________________________
> slicer-users mailing list
> [hidden email]
> http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
> To unsubscribe: send email to
> [hidden email] with unsubscribe as the
> subject

_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject
Reply | Threaded
Open this post in threaded view
|

Re: How to get most accurate 3D volume?

pieper
Administrator
Hi Jiajia -

1) Another thing you might try is the Modules->Filtering->Resample
Scalar Volume to interpolate new slices between the 5mm slices so you
will get a bit better results.

2) You should use the windowed sinc in general - really you need to pick
the number of iterations that look good to you (faithful to the
underlying anatomy).

3) No, that only matters if you have multiple models and you want them
to have a watertight boundary.

-Steve

On J/6/10 11:45 AM, [hidden email] wrote:

> I'm not sure what you are looking for in terms of improved accuracy, but
> have a look at this newer algorithm, and give it a try:
>
>
> http://wiki.na-mic.org/Wiki/index.php/RobustStatisticsSegmentation
>
> It needs to be run in 3.5, and it does give nice results; it might help
> remedy some of the smoothing issues you're are encountering.
>
> Stuart
>
>
>
>> Hi all,
>>
>> I have created several 3D models from Labelmaps on MRI scans. The muscle I
>> am tracing is small and most accurate in the coronal view, however when
>> labelmap is complete the model created is poor partly because of the 5mm
>> gap
>> between MRIs and partly because the lofting is incorrect. I therefore have
>> to make 'fat' tracings in order to make the model look and loft correctly.
>> I
>> also have to use Laplacian smoothing function.
>>
>> Questions:
>> 1) If I trace accurately the model is full of holes and there is no
>> correction function for this is there?
>> 2) Which is the most accurate volume for the model (based on the 'fat'
>> labelmap tracing) ie one at 5,10,20,30,40 Laplacian or Sinc? And what is
>> the
>> difference between Laplacian and Sinc (filter type) in 3D slicer?
>> 3) In a single element model should I tick the joint smoothing box ( as
>> this
>> does make a differnce to the appearance ) or does this make it even less
>> accurate?
>>
>> Thanks,
>>
>> Jiajia
>> _______________________________________________
>> slicer-users mailing list
>> [hidden email]
>> http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
>> To unsubscribe: send email to
>> [hidden email] with unsubscribe as the
>> subject
>
> _______________________________________________
> slicer-users mailing list
> [hidden email]
> http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
> To unsubscribe: send email to [hidden email] with unsubscribe as the subject
_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject
Reply | Threaded
Open this post in threaded view
|

Scale Issues?

Justin Adams
Was hoping that there is a ready solution to the following issue that I have run across:  

I am manually segmenting cranial spaces with dry skulls to assess volumes.  However, the volumes that I reconstruct (say endocranial space or a sinus space) and measure (in mm3) in Slicer are significantly different from expected  (based on published volumes).  For example, the frontal sinuses of one of my crania come out as around 1,600 mm3; but published values suggest that the volume should be around 4,000 mm3.  The original skulls were scanned at a 3.3mm slice width, but with 1.6mm spacing between slices to produce 1.6mm-thick slices.  The original DICOM files were loaded into ImageJ, 'black' space was cropped to reduce file size, then converted to .HDR to be loaded into Slicer for analysis.

Is it possible that my CT scans have gotten rescaled somehow and there is a correction factor I need to apply to get an accurate volume measurement?  Any assistance or advice would be appreciated, I've tried just about everything I can think of...

Cheers

Justin Adams

Department of Biomedical Sciences
Grand Valley State University
Allendale, MI 49401
[hidden email]
and

School of Anatomical Sciences
University of the Witwatersrand
7 York Road
Parktown
Johannesburg, RSA  2193


_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject
Reply | Threaded
Open this post in threaded view
|

Re: Scale Issues?

Kikinis, Ron,M.D.
The info tab in the volume module will show you what dimensions Slicer
was using for the voxels. I believe that you can also manually adjust
them to your liking.

On 1/8/10 10:55 AM, Justin Adams wrote:

> Was hoping that there is a ready solution to the following issue that I have run across:
>
> I am manually segmenting cranial spaces with dry skulls to assess volumes.  However, the volumes that I reconstruct (say endocranial space or a sinus space) and measure (in mm3) in Slicer are significantly different from expected  (based on published volumes).  For example, the frontal sinuses of one of my crania come out as around 1,600 mm3; but published values suggest that the volume should be around 4,000 mm3.  The original skulls were scanned at a 3.3mm slice width, but with 1.6mm spacing between slices to produce 1.6mm-thick slices.  The original DICOM files were loaded into ImageJ, 'black' space was cropped to reduce file size, then converted to .HDR to be loaded into Slicer for analysis.
>
> Is it possible that my CT scans have gotten rescaled somehow and there is a correction factor I need to apply to get an accurate volume measurement?  Any assistance or advice would be appreciated, I've tried just about everything I can think of...
>
> Cheers
>
> Justin Adams
>
> Department of Biomedical Sciences
> Grand Valley State University
> Allendale, MI 49401
> [hidden email]
> and
>
> School of Anatomical Sciences
> University of the Witwatersrand
> 7 York Road
> Parktown
> Johannesburg, RSA  2193
>
>
> _______________________________________________
> slicer-users mailing list
> [hidden email]
> http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
> To unsubscribe: send email to [hidden email] with unsubscribe as the subject

--
Ron Kikinis, M.D.,
Professor of Radiology, Harvard Medical School
Director, Surgical Planning Laboratory
http://www.spl.harvard.edu/~kikinis
_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject
Reply | Threaded
Open this post in threaded view
|

Re: Scale Issues?

swallace
In reply to this post by Justin Adams
Hiya Justin-

A couple of questions:

1. What slicer version are you using?
2. What platform (mac, windows, linux)?
3. Have you tried loading the images in their native dicom format?


Stuart

> Was hoping that there is a ready solution to the following issue that I
> have run across:
>
> I am manually segmenting cranial spaces with dry skulls to assess volumes.
>  However, the volumes that I reconstruct (say endocranial space or a sinus
> space) and measure (in mm3) in Slicer are significantly different from
> expected  (based on published volumes).  For example, the frontal sinuses
> of one of my crania come out as around 1,600 mm3; but published values
> suggest that the volume should be around 4,000 mm3.  The original skulls
> were scanned at a 3.3mm slice width, but with 1.6mm spacing between slices
> to produce 1.6mm-thick slices.  The original DICOM files were loaded into
> ImageJ, 'black' space was cropped to reduce file size, then converted to
> .HDR to be loaded into Slicer for analysis.
>
> Is it possible that my CT scans have gotten rescaled somehow and there is
> a correction factor I need to apply to get an accurate volume measurement?
>  Any assistance or advice would be appreciated, I've tried just about
> everything I can think of...
>
> Cheers
>
> Justin Adams
>
> Department of Biomedical Sciences
> Grand Valley State University
> Allendale, MI 49401
> [hidden email]
> and
>
> School of Anatomical Sciences
> University of the Witwatersrand
> 7 York Road
> Parktown
> Johannesburg, RSA  2193
>
>
>
>
>

_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject
Reply | Threaded
Open this post in threaded view
|

Re: Scale Issues?

Isaiah Norton
In reply to this post by Kikinis, Ron,M.D.
As a sanity check you could use the fiducial distance measurement (just
below the list of points in fiducials module) to compare Slicer
distances to some (presumably known) real-world distances for exterior
points.

-I



On Fri, 2010-01-08 at 12:37 -0500, Ron Kikinis wrote:

> The info tab in the volume module will show you what dimensions Slicer
> was using for the voxels. I believe that you can also manually adjust
> them to your liking.
>
> On 1/8/10 10:55 AM, Justin Adams wrote:
> > Was hoping that there is a ready solution to the following issue that I have run across:
> >
> > I am manually segmenting cranial spaces with dry skulls to assess volumes.  However, the volumes that I reconstruct (say endocranial space or a sinus space) and measure (in mm3) in Slicer are significantly different from expected  (based on published volumes).  For example, the frontal sinuses of one of my crania come out as around 1,600 mm3; but published values suggest that the volume should be around 4,000 mm3.  The original skulls were scanned at a 3.3mm slice width, but with 1.6mm spacing between slices to produce 1.6mm-thick slices.  The original DICOM files were loaded into ImageJ, 'black' space was cropped to reduce file size, then converted to .HDR to be loaded into Slicer for analysis.
> >
> > Is it possible that my CT scans have gotten rescaled somehow and there is a correction factor I need to apply to get an accurate volume measurement?  Any assistance or advice would be appreciated, I've tried just about everything I can think of...
> >
> > Cheers
> >
> > Justin Adams
> >
> > Department of Biomedical Sciences
> > Grand Valley State University
> > Allendale, MI 49401
> > [hidden email]
> > and
> >
> > School of Anatomical Sciences
> > University of the Witwatersrand
> > 7 York Road
> > Parktown
> > Johannesburg, RSA  2193
> >
> >
> > _______________________________________________
> > slicer-users mailing list
> > [hidden email]
> > http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
> > To unsubscribe: send email to [hidden email] with unsubscribe as the subject
>

_______________________________________________
slicer-users mailing list
[hidden email]
http://massmail.spl.harvard.edu/mailman/listinfo/slicer-users
To unsubscribe: send email to [hidden email] with unsubscribe as the subject