<table cellspacing="0" cellpadding="0" border="0" ><tr><td valign="top" style="font: inherit;"><span class="Apple-style-span" style="font-family: 'Times New Roman'; font-size: medium; "><pre>Hi John,</pre><pre><br></pre><pre>I don't think performing the diffusion filter twice would be the </pre><pre><br></pre><pre>source of the problem unless if its parameter settings would be </pre><pre><br></pre><pre>extremely out of proportion.</pre><pre><br></pre><pre>For better convergence it is best to have a good initial estimation</pre><pre><br></pre><pre>of the level set as a seed image. Larger number of iterations alone</pre><pre><br></pre><pre>does not guarantee the convergence.</pre><pre><br></pre><pre>Basically using the diffusion filter improves the quality of the</pre><pre><br></pre><pre>segmentation, however over-diffusing the images can cause loss of </pre><pre><br></pre><pre>details. </pre><pre><br></pre><pre>You can use the
resulting image (SpeedImage.mha) to determine that </pre><pre><br></pre><pre>to what extent you want to diffuse the input image. Increasing the </pre><pre><br></pre><pre>number of diffusion iterations and lower values for the conductance</pre><pre><br></pre><pre>will give a more diffused image.</pre><pre><br></pre><pre>Propagation weight determines the relative amount emphasis on </pre><pre><br></pre><pre>the propagation speed and with higher values results in narrower </pre><pre><br></pre><pre>surfaces, opposing to the curvature weight (curvature scaling) which </pre><pre><br></pre><pre>with higher values gains smoother surfaces.</pre><pre><br></pre><pre>The isovalue I think is best to be assigned as halfway between the</pre><pre><br></pre><pre>maximum and minimum values in the seed image. </pre><pre><br></pre><pre>Also, Premature elimination might be the result of the low maximum </pre><pre><br></pre><pre>allowed
RMS.</pre><pre><br></pre><pre>As for number of iterations for both diffusion and segmentation after</pre><pre><br></pre><pre>couple of tries and assigning different values you should be able to</pre><pre><br></pre><pre>find the best fit for your application.</pre><pre><br></pre><pre>Best regards,</pre><pre><br></pre><pre>Dawood</pre><pre><br></pre><pre><br></pre><pre>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<</pre><pre><br></pre><pre><span class="Apple-style-span" style="font-family: 'Times New Roman'; white-space: normal;
"><pre>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<</pre></span></pre><pre><br></pre><pre><br></pre><pre>Hi again,
><i>
</i>><i> I have made some progress on my own by randomly selecting an isovalue of
</i>><i> 0.5. Now my level set segmentation is no longer pitted, but it has the same
</i>><i> volume in mm^3 as the initially inputted fuzzy connectedness segmentation.
</i>><i> It only performed one iteration and the RMS change was 0.0014962. The
</i>><i> maximum RMS error was set to 0.002 and the maximum number of iterations was
</i>><i> set to 100. The number of diffusion iterations was set to 10, the diffusion
</i>><i> conductance was set to 2.0, the propagation weight was set to 1.0, the
</i>><i> initial model isovalue was set to 0.5, and the maximum number of iterations
</i>><i> was set to 100.
</i>><i>
</i>><i> Any further advice is welcome.
</i>><i>
</i>><i>
</i>Thanks,
><i>
</i>
><i> John
</i>><i>
</i>
><i> On Thu, Feb 3, 2011 at 12:11 PM, John Drozd <<a href="http://www.itk.org/mailman/listinfo/insight-users">john.drozd at gmail.com</a>> wrote:
</i>><i>
</i>>><i> Hello,
</i>>><i> I have performed a reasonable fuzzy connectedness segmentation (that I
</i>>><i> modified to use multiple seeds) of lateral human brain ventricles in an MRI
</i>>><i> brain image of a mildly cognitive impaired person that I obtained from the
</i>>><i> ADNI database.
</i>>><i> Prior to performing the fuzzy connectedness segmentation, I smoothed the
</i>>><i> noise and enhanced the edges using an Anisotropic Diffusion filter.
</i>>><i> I am trying to use this fuzzy connectedness segmentation as my initial
</i>>><i> level set for a level set segmentation but when performing a Laplacian Level
</i>>><i> Set segmentation, I get a pitted volume.
</i>>><i> I think the following are my problems.
</i>>><i> 1) When I load the label map from the fuzzy connectedness segmentation
</i>>><i> into 3D Slicer to view it, I get Fg: None and Bg: 1 for the segmented
</i>>><i> volume, and Fg: None and Bg: 0 for the black area that surrounds the
</i>>><i> ventricles. I think that I am setting my foreground and background values
</i>>><i> incorrectly. I feel this is a problem because I need to specify an isovalue
</i>>><i> midway between the foreground and background values. I can send you my
</i>>><i> fuzzy connectedness and Laplacian level set segmentation code if this would
</i>>><i> help you be able to help me. Also, the laplacian level set segmentation
</i>>><i> performs anisotropic diffusion. Because I perform this prior to my fuzzy
</i>>><i> connectedness segmentation, I am essentially performing anisotropic
</i>>><i> diffusion twice. Could this be a problem?
</i>>><i> 2) Also, how do you run the level set segmentation to go to convergence.
</i>>><i> Do you just input a large maximum number of iterations?
</i>>><i> 3) Also how can I use the outputted "speedImage.mha" figure to help me
</i>>><i> tune my parameters for "DiffusionIterations", "DiffusionConductance",
</i>>><i> "PropagationWeight", "InitialModelIsovalue" and "MaximumIterations". I can
</i>>><i> send you the speed image as well.
</i>>><i> 4) Also, the level set segmentation only performs 1 iteration eventhough I
</i>>><i> specified a maximum 100 iterations. Should I use a smaller RMS error than
</i>>><i> 0.002?
</i>>><i>
</i>>><i> Any help would be appreciated.
</i>>><i>
</i>>><i> Thanks for your time,
</i>>><i>
</i>>><i> John
</i>>><i>
</i>>><i> --
</i>>><i> John Drozd
</i>>><i> Post-Doctoral Fellow, Robarts Research Institute
</i>>><i> The University of Western Ontario
</i>>><i> London, ON, Canada
</i>>></pre></span></td></tr></table><br>