The beam verification set consists of a series of dose distributions which goes beyond the beam characterisation set, because it shall be used to test the algorithm. It contains more depth dose and profile data than the curves needed for the configuration (more combinations of field size, energy, SSD and so on). The beam verification set also contains oblique beams and is not restricted to the water phantom. Heterogeneities, non-flat surfaces, missing lateral scatter etc. are also checked. The only question is: what verification checks belong to the standard program every department should follow, and what checks should be left to the vendor or academic institutions who usually have more resources for such a task? The paper of Ding (2006) covers many aspects of dose behind heterogeneities: if one believes the findings in the paper (and I see no reason why I should not do it), one can skip most of the heterogeneity checks, because the algorithm was verified in this respect by someone else. There are two groups of Setups checked by us:
The Calculation Options in Eclipse are chosen with an accuracy high enough to test the algorithm with reasonable calculation time:
Grids down to 1mm are possible, but the calculation time for a large number of plans would be quite high. |
The Calculation Queue |
The calculation queue is a great feature of Eclipse (not only with eMC). In the next screenshot, a calculated dose distribution is displayed in Eclipse. Three more plans are currently being calculated in the background. Calculations run locally on the Eclipse workstation or via network on one of the other calculation servers. |
I always distribute the plans on different calculation servers to distribute the computing load. The job can be sent to the queue via Shift+F5, the queue can be looked at via Ctrl+F5. In the current version of Eclipse, the server on which a certain job is running cannot be seen in the Calculation Queue Manager. This would be a nice feature in a future release (automatic grid computing - the network is the computer - would be even better). When the job is finished and loaded, the Log displays information about the calculation time and the number of particle histories. The calculation was done on s10iroec004 (calculation server number 4) with eMC 7.5.22: |
Until the desired accuracy of 1% was reached, 27Mio. particles had to be
processed. The net calculation
speed therefore was 25.000 histories per second. |