<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <p>Hey,<br>
      <br>
      Thanks for the useful hints and the literature.</p>
    <p> Can you explain a little bit more how the uncertainty image is
      created and how it is best to understand? What tell me the numbers
      or values in the image? I think this is important to understand,
      in order to choose the number of photons correctly. <br>
    </p>
    <p>Yes, that's what I am doing already. I scaled the detectors
      resolution and the voxelized volume down by a factor of 4 and
      consequently simulate the scatter images.<br>
    </p>
    <p>One more question concerning the engine seed. Do you choose for
      every projection a new seed by changing the order of the numbers
      (123456) or is that not necessary?</p>
    <p>Cheers,</p>
    <p>Nico<br>
    </p>
    <div class="moz-cite-prefix">On 02/09/2017 01:42 PM, Simon Rit
      wrote:<br>
    </div>
    <blockquote
cite="mid:CAF0oig3ykop+JyiF5fVw8hG2m4hmgFt6HfS2isMZcYUQuHK7YQ@mail.gmail.com"
      type="cite">
      <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
      <div dir="ltr">
        <div>
          <div>
            <div>
              <div>Hi,<br>
              </div>
              Python is probably easier indeed.<br>
            </div>
            The stochastic part is for scatter and secondary radiations
            (compton, rayleigh and fluorescence). FFD uses a low
            statistics Monte Carlo simulation (therefore stochastic) and
            combines it with a deterministic calculation. Useful
            unordered references to understand the technique:<br>
            <a moz-do-not-send="true"
              href="http://dx.doi.org/10.1088/0031-9155/54/12/016">dx.doi.org/10.1088/0031-9155/54/12/016</a><br>
            <a moz-do-not-send="true"
              href="http://dx.doi.org/10.1109/TMI.2004.825600">dx.doi.org/10.1109/TMI.2004.825600</a><br>
            <a moz-do-not-send="true"
              href="http://doi.org/10.1109/TNS.2005.858223">doi.org/10.1109/TNS.2005.858223</a><br>
          </div>
          1000 is not sufficient, I typically use 10^5 photons at least
          for one projection. The best is to record the uncertainty
          image to have an estimate of the precision of your Monte Carlo
          simulation (using the <b>enableUncertaintySecondary</b>
          option). You'll probably want to limit the number of pixels of
          your projection to accelerate the computation of your scatter
          images. I typically use finer lattices for primary images than
          for scatter images.<br>
        </div>
        Simon<br>
        <div>
          <div>
            <div><br>
            </div>
          </div>
        </div>
      </div>
      <div class="gmail_extra"><br>
        <div class="gmail_quote">On Thu, Feb 9, 2017 at 1:11 PM,
          Triltsch, Nicolas <span dir="ltr"><<a
              moz-do-not-send="true"
              href="mailto:nicolas.triltsch@tum.de" target="_blank">nicolas.triltsch@tum.de</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div bgcolor="#FFFFFF" text="#000000">
              <p>Hey Simon,</p>
              <p>Thanks for your always very helpful answers. For your
                last point, I found out a little work around. I used an
                alias of the form "Gate [rot_angle,$(angle*i)][run_id,<wbr>$i]
                mymacro.mac" combined with a for loop with parameter i.
                In the macro I named the files in the output folder
                "output_files{i}". I can highly recommend not to use
                bash scripts to execute the Gate commands which where
                created in a for loop style. Defining variables
                (especially floats, etc.) is quite handy. It's better to
                use a python script and execute the Gate macros in a so
                called subprocess.call(), multiprocessing module is
                required here.  <br>
              </p>
              <p>Some more questions popped up my mind while I was
                reading your email.</p>
              <p>- I think you didn't understood me correctly. I used
                1000 photons per projection, not in total. I am NOT only
                interested in the primary image, but also in the images
                compton.mha and rayleigh.mha. Where in the calculation
                comes the NOT deterministic part? And do you have any
                experience how many photons are necessary for a
                trustworthy result?<br>
              </p>
              <p>Thanks in advance,</p>
              <p>Nico<br>
              </p>
              <div>
                <div class="h5">
                  <div class="m_-5837947688201281229moz-cite-prefix">On
                    02/07/2017 05:37 PM, Simon Rit wrote:<br>
                  </div>
                  <blockquote type="cite">
                    <div dir="ltr">
                      <div>
                        <div>
                          <div>
                            <div>
                              <div>
                                <div>
                                  <div>Hi Nicolas,<br>
                                  </div>
                                  Good to see that ffda is used. To
                                  answer your questions:<br>
                                </div>
                                - yes, there is an "intrinsic
                                parallelization". The number of threads
                                is set by the environment variable
                                ITK_GLOBAL_DEFAULT_NUMBER_OF_T<wbr>HREADS.
                                If you haven't set it, it will use all
                                your cores. The Monte Carlo part is
                                still single-threaded, but the ray
                                casting is multi-threaded using RTK
                                (based on ITK).<br>
                              </div>
                              - you use 1000 photons. I guess you're
                              only interested in the primary image? In
                              this case, 1 photon per projection is
                              enough since the primary part is
                              deterministic.<br>
                            </div>
                            - you can use the printf format to set the
                            run id in the file primary name (see line
                            842 of <a moz-do-not-send="true"
href="https://github.com/OpenGATE/Gate/blob/develop/source/digits_hits/src/GateFixedForcedDetectionActor.cc#L842"
                              target="_blank">GateFixedForcedDetectionActor.<wbr>cc</a>):<br>
                            /gate/actor/ffda/primaryFilena<wbr>me   
                            output/primary%0d.mha<br>
                          </div>
                          - for further parallelization, I would suggest
                          to run Gate on several machines, each machine
                          starting at a different angle and over an
                          angle range which would be limited. This
                          requires some specific dev and careful
                          handling of all the outputs (they all start
                          with a runid of 0, so you will need to rename
                          the outputs).<br>
                        </div>
                        I hope this helps.<br>
                      </div>
                      Simon<br>
                      <div>
                        <div>
                          <div>
                            <div>
                              <div>
                                <div>
                                  <div><br>
                                  </div>
                                </div>
                              </div>
                            </div>
                          </div>
                        </div>
                      </div>
                    </div>
                    <div class="gmail_extra"><br>
                      <div class="gmail_quote">On Tue, Feb 7, 2017 at
                        10:58 AM, Triltsch, Nicolas <span dir="ltr"><<a
                            moz-do-not-send="true"
                            href="mailto:nicolas.triltsch@tum.de"
                            target="_blank">nicolas.triltsch@tum.de</a>></span>
                        wrote:<br>
                        <blockquote class="gmail_quote" style="margin:0
                          0 0 .8ex;border-left:1px #ccc
                          solid;padding-left:1ex">
                          <div bgcolor="#FFFFFF" text="#000000">
                            <p>Hello Gate community,</p>
                            <p>I am using the fixed forced detection
                              actor (ffda) and I try to run a full CT
                              simulation with 1201 projections. My first
                              question aims to the possibilities of
                              parallelization. I noticed that if I run a
                              single projection, all 4 cores of my local
                              computer are running at almost 100%. Is
                              there already some intrinsic
                              parallelization step when using the ffda
                              and what further parallelization steps are
                              possible to speed up the simulation for 1
                              projection? If it helps I use a voxelized
                              phantom, cone beam setup, a xray spectrum
                              histogram, integrating detector and 1000
                              Photons.<br>
                            </p>
                            <p>My second question is how to save several
                              .mha images in the output folder when
                              simulating all 1201 projections. Still, I
                              am using the ffda actor and with the
                              command "<i>/gate/actor/ffda/primaryFilen<wbr>ame   
                                output/primary.mha" </i>the primary
                              image gets overwritten for each
                              projection. How can I save different
                              primary images for each projection?<br>
                            </p>
                            <p>Any help is appreciated!<br>
                            </p>
                            <p>Nico<br>
                            </p>
                            <pre class="m_-5837947688201281229m_4700849550904366916moz-signature" cols="72">-- 
B.Sc. Nicolas Triltsch
Masterand

Technische Universität München
Physik-Department
Lehrstuhl für Biomedizinische Physik E17

James-Franck-Straße 1
85748 Garching b. München

Tel: <a moz-do-not-send="true" href="tel:+49%2089%2028912591" value="+498928912591" target="_blank">+49 89 289 12591</a>

<a moz-do-not-send="true" class="m_-5837947688201281229m_4700849550904366916moz-txt-link-abbreviated" href="mailto:nicolas.triltsch@tum.de" target="_blank">nicolas.triltsch@tum.de</a>
<a moz-do-not-send="true" class="m_-5837947688201281229m_4700849550904366916moz-txt-link-abbreviated" href="http://www.e17.ph.tum.de" target="_blank">www.e17.ph.tum.de</a></pre>
                          </div>
                          <br>
                          ______________________________<wbr>_________________<br>
                          Gate-users mailing list<br>
                          <a moz-do-not-send="true"
                            href="mailto:Gate-users@lists.opengatecollaboration.org"
                            target="_blank">Gate-users@lists.opengatecolla<wbr>boration.org</a><br>
                          <a moz-do-not-send="true"
href="http://lists.opengatecollaboration.org/mailman/listinfo/gate-users"
                            rel="noreferrer" target="_blank">http://lists.opengatecollabora<wbr>tion.org/mailman/listinfo/<wbr>gate-users</a><br>
                        </blockquote>
                      </div>
                      <br>
                    </div>
                  </blockquote>
                  <br>
                  <pre class="m_-5837947688201281229moz-signature" cols="72">-- 
B.Sc. Nicolas Triltsch
Masterand

Technische Universität München
Physik-Department
Lehrstuhl für Biomedizinische Physik E17

James-Franck-Straße 1
85748 Garching b. München

Tel: <a moz-do-not-send="true" href="tel:+49%2089%2028912591" value="+498928912591" target="_blank">+49 89 289 12591</a>

<a moz-do-not-send="true" class="m_-5837947688201281229moz-txt-link-abbreviated" href="mailto:nicolas.triltsch@tum.de" target="_blank">nicolas.triltsch@tum.de</a>
<a moz-do-not-send="true" class="m_-5837947688201281229moz-txt-link-abbreviated" href="http://www.e17.ph.tum.de" target="_blank">www.e17.ph.tum.de</a></pre>
                </div>
              </div>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
    <pre class="moz-signature" cols="72">-- 
B.Sc. Nicolas Triltsch
Masterand

Technische Universität München
Physik-Department
Lehrstuhl für Biomedizinische Physik E17

James-Franck-Straße 1
85748 Garching b. München

Tel: +49 89 289 12591

<a class="moz-txt-link-abbreviated" href="mailto:nicolas.triltsch@tum.de">nicolas.triltsch@tum.de</a>
<a class="moz-txt-link-abbreviated" href="http://www.e17.ph.tum.de">www.e17.ph.tum.de</a></pre>
  </body>
</html>