<div dir="ltr"><div><div><div>Thanks very much for the reply.<br><br></div>the example from here <a href="http://itk.org/ITKExamples/src/Nonunit/Review/SegmentBloodVesselsWithMultiScaleHessianBasedMeasure/Documentation.html" target="_blank">http://itk.org/ITKExamples/src/Nonunit/Review/SegmentBloodVesselsWithMultiScaleHessianBasedMeasure/Documentation.html</a> is complete, and works great in the 2D case. When I tried to extend the same filter for the 3D case however I end up with an empty volume. I looked through both of the following<br>
ITK/Modules/Nonunit/Review/test/itkMultiScaleHessianBasedMeasureImageFilterTest.cxx<br><div>
ITK/Modules/Nonunit/Review/test/itkHessianToObjectnessMeasureImageFilterTest.cxx</div>To try to get a better idea, but the examples for these are also in 2D also. <br></div>Reading the original paper in the submission of the filters to ITK i find an example in 3d but it seems the filter has changed quite a bit since 2007.<br>
<br></div>For anyone coming across this thread in the future, the solution for me was to set objectnessFilter->SetScaleObjectnessMeasure( true ); From the documentation "Toggle scaling the objectness measure with the magnitude of the largest absolute eigenvalue " I am reading further to figure out why this is necessary for the 3D case but now I am able to generate the segmented vessels using the multiscale Hessian filter.<br>
<div>
<br>Many thanks for your help,<br>Jesse<br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Thu, Aug 22, 2013 at 6:39 PM, Matt McCormick <span dir="ltr"><<a href="mailto:matt.mccormick@kitware.com" target="_blank">matt.mccormick@kitware.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">On Thu, Aug 22, 2013 at 3:23 PM, Kevin H. Hobbs <<a href="mailto:hobbsk@ohio.edu">hobbsk@ohio.edu</a>> wrote:<br>
> On 08/22/2013 10:35 AM, Jesse Ross-Jones wrote:<br>
>> Hi,<br>
>> Thanks for the quick reply.<br>
>> Reading the documentation on the MultiScaleHessian, "The Hessian-based<br>
>> measure is computed from the Hessian image at each scale level and the<br>
>> best response is selected." Does this mean that the final image produced<br>
>> is the result at only one scale producing the best response or a mixture<br>
>> of the results at different scales and each pixel is takes the value<br>
>> from the hessian image giving the best response?<br>
><br>
> The output images have at each pixel the greatest response and the scale<br>
> that produced it.<br>
><br>
> I just took a look at :<br>
><br>
> ITK/Modules/Nonunit/Review/include/itkMultiScaleHessianBasedMeasureImageFilter.hxx<br>
><br>
> to make sure.<br>
><br>
> The decision is made in the function UpdateMaximumResponse.<br>
><br>
<br>
</div></div>Yes, looking at the source code always has the most detailed,<br>
up-to-date explanation :-).<br>
<br>
This Insight Journal article for this class (referenced earlier) is<br>
informative, too.<br>
<br>
Here is an example:<br>
<br>
<a href="http://itk.org/ITKExamples/src/Nonunit/Review/SegmentBloodVesselsWithMultiScaleHessianBasedMeasure/Documentation.html" target="_blank">http://itk.org/ITKExamples/src/Nonunit/Review/SegmentBloodVesselsWithMultiScaleHessianBasedMeasure/Documentation.html</a><br>
<br>
HTH,<br>
Matt<br>
</blockquote></div><br></div>