[Insight-users] Laplacian level set filter
Radhika Sivaramakrishna
radhika.sivaramakrishna at synarc.com
Fri, 30 Jan 2004 08:25:35 -0800
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
------_=_NextPart_001_01C3E74D.B09D29B6
Content-Type: text/plain
Hi Luis and Josh,
Thanks for your suggestions. I did a number of things though the results
don't seem much improved.
For my example, I kept the following fixed:
# iterations for diffusion = 10,
time step = 0.125
curvature scaling for laplacian = 1.0
isosurface value = 127.5.
# iterations for laplacian = 50
Maximum rms error = 0.002
I varied conductance parameter from 0.5 to 2.0 in steps of 0.5.
I also varied propagation scaling from 0.5 to 1.5 in steps of 0.5
I varied Advection scaling from 30 to 90 in steps of 30.
However, the results in each of these cases is not much improved. The front
does not appear to have advanced much.
1) I looked visually at the diffusion image. When I change the conductance
parameter from 0.5 to 2.0 (keeping other parameters constant), this tends to
make the image smoother (to the point of washing out all the details inside)
however, the edges that I am interested in appear well preserved.
2) I looked at several of the speed images and they do appear to have zero
crossings close to edges I am interested in. However, the speed image
appears very rough so there are several zero crossings elsewhere too.
However, I saw that when the conductance parameter is very high, the details
inside the speed image appear washed out, but there are still zero crossings
at the boundaries that I am trying to get.
3) In all cases, the maximum number of iterations are used up. The RMS
change (result of GetRMSChange is something like 0.00685403 to 0.00907139)
with conductance parameter influencing it the most. Advection scaling
appears to have no influence on this.
Radhika
-----Original Message-----
From: Luis Ibanez [mailto:luis.ibanez at kitware.com]
Sent: Thursday, January 29, 2004 9:11 AM
To: Radhika Sivaramakrishna
Cc: ITK
Subject: Re: [Insight-users] Laplacian level set filter
Hi Radhika,
Did the LaplacianSegmentationLevelSet
filter use all the iterations that you set ?
That its, did it reached the target RMS error
before using all the iterations ?
If not,
your parameters are probably too conservative.
If you need to accelerate the propagation of
the front you could try using values as high
as 10 to 100 for the Translation scaling as
well as the Advection scaling.
Please also check visually the result of
smoothing the MRI image before you pass it
as input to the LevelSet filter.
Regards,
Luis
--------------------------------
Radhika Sivaramakrishna wrote:
> Hi Luis,
>
> I was trying to use the Laplacian level set (with anisotropic diffiusion
> as a preprocessing step) to improve an already available initial mask.
> These are 3D brain T2-weighted images.
>
>
>
> I set the # of iterations for diffusion as 10
>
> Time step for diffusion = 0.125
>
> Conductance parameter for diffusion = 2.0
>
>
>
> Curvature scaling = 1.0
>
> Propagation Scaling for Laplacian = 1.0
>
> # iterations for Laplacian = 10
>
> Isosurface value = 127.5
>
> Maximum rms error for laplacian = 0.002
>
>
>
> The initial mask was quite approximate but the solution was a good
> starting point. However, after running with above settings, I found very
> marginal improvement.
>
> I was playing with some of the parameters but did not see significant
> improvement.
>
>
>
> So my question was, for a 3D MRI T2-weighted image,
>
>
>
> 1) what are the recommended values for the above parameters
>
> 2) What is the range of reasonable values to "play with" for each
> of these parameters?
>
> 3) What are the parameters for the Laplacian scaling that make the
> growing surface move faster? It seems to me that the advancement at each
> iteration is very less.
>
>
>
> Thanks
>
> Radhika
>
>
>
>
>
>
>
>
>
>
>
> -----------------------------------------------------
>
> Confidentiality Notice.
>
> This email message is for the sole use of the intended recipient(s) and
> may contain confidential and privileged information. Any unauthorized
> review, use, disclosure or distribution is prohibited. If you are not
> the intended recipient, please contact the sender by reply email and
> destroy all copies of the original message. If you are the intended
> recipient, please be advised that the content of this message is subject
> to access, review and disclosure by the sender's Email System
Administrator.
>
-----------------------------------------------------
Confidentiality Notice.
This email message is for the sole use of the intended recipient(s) and may
contain confidential and privileged information. Any unauthorized review,
use, disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message. If you are the intended recipient, please be
advised that the content of this message is subject to access, review and
disclosure by the sender's Email System Administrator.
------_=_NextPart_001_01C3E74D.B09D29B6
Content-Type: text/html
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Dus-ascii">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
5.5.2656.87">
<TITLE>RE: [Insight-users] Laplacian level set filter</TITLE>
</HEAD>
<BODY>
<P><FONT SIZE=3D2>Hi Luis and Josh,</FONT>
<BR><FONT SIZE=3D2>Thanks for your suggestions. I did a number of =
things though the results don't seem much improved. </FONT>
<BR><FONT SIZE=3D2>For my example, I kept the following fixed:</FONT>
<BR><FONT SIZE=3D2># iterations for diffusion =3D 10, </FONT>
<BR><FONT SIZE=3D2>time step =3D 0.125</FONT>
<BR><FONT SIZE=3D2>curvature scaling for laplacian =3D 1.0</FONT>
<BR><FONT SIZE=3D2>isosurface value =3D 127.5.</FONT>
<BR><FONT SIZE=3D2># iterations for laplacian =3D 50</FONT>
<BR><FONT SIZE=3D2>Maximum rms error =3D 0.002</FONT>
</P>
<P><FONT SIZE=3D2>I varied conductance parameter from 0.5 to 2.0 in =
steps of 0.5. </FONT>
<BR><FONT SIZE=3D2>I also varied propagation scaling from 0.5 to 1.5 in =
steps of 0.5</FONT>
<BR><FONT SIZE=3D2>I varied Advection scaling from 30 to 90 in steps of =
30.</FONT>
</P>
<P><FONT SIZE=3D2>However, the results in each of these cases is not =
much improved. The front does not appear to have advanced much. </FONT>
</P>
<P><FONT SIZE=3D2>1) I looked visually at the diffusion image. When I =
change the conductance parameter from 0.5 to 2.0 (keeping other =
parameters constant), this tends to make the image smoother (to the =
point of washing out all the details inside) however, the edges that I =
am interested in appear well preserved. </FONT></P>
<P><FONT SIZE=3D2>2) I looked at several of the speed images and they =
do appear to have zero crossings close to edges I am interested in. =
However, the speed image appears very rough so there are several zero =
crossings elsewhere too. However, I saw that when the conductance =
parameter is very high, the details inside the speed image appear =
washed out, but there are still zero crossings at the boundaries that I =
am trying to get.</FONT></P>
<P><FONT SIZE=3D2>3) In all cases, the maximum number of iterations are =
used up. The RMS change (result of GetRMSChange is something like =
0.00685403 to 0.00907139) with conductance parameter influencing it the =
most. Advection scaling appears to have no influence on =
this.</FONT></P>
<P><FONT SIZE=3D2>Radhika</FONT>
</P>
<BR>
<BR>
<P><FONT SIZE=3D2>-----Original Message-----</FONT>
<BR><FONT SIZE=3D2>From: Luis Ibanez [<A =
HREF=3D"mailto:luis.ibanez at kitware.com">mailto:luis.ibanez at kitware.com</=
A>] </FONT>
<BR><FONT SIZE=3D2>Sent: Thursday, January 29, 2004 9:11 AM</FONT>
<BR><FONT SIZE=3D2>To: Radhika Sivaramakrishna</FONT>
<BR><FONT SIZE=3D2>Cc: ITK</FONT>
<BR><FONT SIZE=3D2>Subject: Re: [Insight-users] Laplacian level set =
filter</FONT>
</P>
<BR>
<P><FONT SIZE=3D2>Hi Radhika,</FONT>
</P>
<P><FONT SIZE=3D2>Did the LaplacianSegmentationLevelSet</FONT>
<BR><FONT SIZE=3D2>filter use all the iterations that you set ?</FONT>
</P>
<P><FONT SIZE=3D2>That its, did it reached the target RMS error</FONT>
<BR><FONT SIZE=3D2>before using all the iterations ?</FONT>
</P>
<P><FONT SIZE=3D2>If not,</FONT>
<BR><FONT SIZE=3D2>your parameters are probably too =
conservative.</FONT>
</P>
<P><FONT SIZE=3D2>If you need to accelerate the propagation of</FONT>
<BR><FONT SIZE=3D2>the front you could try using values as high</FONT>
<BR><FONT SIZE=3D2>as 10 to 100 for the Translation scaling as</FONT>
<BR><FONT SIZE=3D2>well as the Advection scaling.</FONT>
</P>
<P><FONT SIZE=3D2>Please also check visually the result of</FONT>
<BR><FONT SIZE=3D2>smoothing the MRI image before you pass it</FONT>
<BR><FONT SIZE=3D2>as input to the LevelSet filter.</FONT>
</P>
<BR>
<P><FONT SIZE=3D2>Regards,</FONT>
</P>
<BR>
<P><FONT SIZE=3D2> Luis</FONT>
</P>
<BR>
<BR>
<P><FONT SIZE=3D2>--------------------------------</FONT>
<BR><FONT SIZE=3D2>Radhika Sivaramakrishna wrote:</FONT>
</P>
<P><FONT SIZE=3D2>> Hi Luis,</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> I was trying to use the Laplacian level set =
(with anisotropic diffiusion </FONT>
<BR><FONT SIZE=3D2>> as a preprocessing step) to improve an already =
available initial mask. </FONT>
<BR><FONT SIZE=3D2>> These are 3D brain T2-weighted images.</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> I set the # of iterations for diffusion as =
10</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Time step for diffusion =3D 0.125</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Conductance parameter for diffusion =3D =
2.0</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Curvature scaling =3D 1.0</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Propagation Scaling for Laplacian =3D =
1.0</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> # iterations for Laplacian =3D 10</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Isosurface value =3D 127.5</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Maximum rms error for laplacian =3D =
0.002</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> The initial mask was quite approximate but the =
solution was a good </FONT>
<BR><FONT SIZE=3D2>> starting point. However, after running with =
above settings, I found very </FONT>
<BR><FONT SIZE=3D2>> marginal improvement.</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> I was playing with some of the parameters but =
did not see significant </FONT>
<BR><FONT SIZE=3D2>> improvement.</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> So my question was, for a 3D MRI T2-weighted =
image,</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> 1) what are =
the recommended values for the above parameters</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> 2) What is =
the range of reasonable values to "play with" for each =
</FONT>
<BR><FONT SIZE=3D2>> of these parameters?</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> 3) What are =
the parameters for the Laplacian scaling that make the </FONT>
<BR><FONT SIZE=3D2>> growing surface move faster? It seems to me =
that the advancement at each </FONT>
<BR><FONT SIZE=3D2>> iteration is very less.</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Thanks</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Radhika</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> =
-----------------------------------------------------</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> Confidentiality Notice.</FONT>
<BR><FONT SIZE=3D2>> </FONT>
<BR><FONT SIZE=3D2>> This email message is for the sole use of the =
intended recipient(s) and </FONT>
<BR><FONT SIZE=3D2>> may contain confidential and privileged =
information. Any unauthorized </FONT>
<BR><FONT SIZE=3D2>> review, use, disclosure or distribution is =
prohibited. If you are not </FONT>
<BR><FONT SIZE=3D2>> the intended recipient, please contact the =
sender by reply email and </FONT>
<BR><FONT SIZE=3D2>> destroy all copies of the original message. If =
you are the intended </FONT>
<BR><FONT SIZE=3D2>> recipient, please be advised that the content =
of this message is subject </FONT>
<BR><FONT SIZE=3D2>> to access, review and disclosure by the =
sender's Email System Administrator.</FONT>
<BR><FONT SIZE=3D2>> </FONT>
</P>
<BR>
<P><FONT =
SIZE=3D2> &nb=
sp; &nb=
sp; &nb=
sp; &nb=
sp; </FONT>
<BR><FONT =
SIZE=3D2>----------------------------------------------------- =
</FONT>
<BR><FONT SIZE=3D2>Confidentiality Notice. </FONT>
<BR><FONT SIZE=3D2>This email message is for the sole use of the =
intended recipient(s) and may contain confidential and privileged =
information. Any unauthorized review, use, disclosure or distribution =
is prohibited. If you are not the intended recipient, please contact =
the sender by reply email and destroy all copies of the original =
message. If you are the intended recipient, please be advised that the =
content of this message is subject to access, review and disclosure by =
the sender's Email System Administrator.</FONT></P>
</BODY>
</HTML>
------_=_NextPart_001_01C3E74D.B09D29B6--