How to get the the output of objective function or Likelyhood function?

General Discussions
1

Dear respected expert,

I have just started using castoR for PET imaging research and would like to ask whether it is possible to output the objective function value or the likelihood probability value during each iteration and sub-step process.

Kind regards,

Bo Li

2

Hi,

You can use the -opti-fom option to print out this.

But, be warned of these two things:

  1. the printed values are computed for the image obtained at the previous update, otherwise it would require another projection at the end of each update
  2. the values are computed only for events considered within the current subset

If you want to compute the complete likelihood of a given image, you can run castor another time with option “-img your_image.hdr” with one iteration and one subset. This option will use the image for initialization so that the computed likelihood values will be done on this image.

Best
Simon

boli November 11 Dear respected expert, I have just started using castoR for PET imaging research and would like to ask whether it is possible to output the objective function value or the likelihood probability value during each iteration and

| boli
November 11 |

  • | - |

Dear respected expert,

I have just started using castoR for PET imaging research and would like to ask whether it is possible to output the objective function value or the likelihood probability value during each iteration and sub-step process.

Kind regards,

Bo Li

3

Dear expert,

Thanks for your help, now I add this -opti-fom option, and I get this error: ***** vOptimizer::CheckParameters() → Computing FOMs in the data-space with list-mode data is not possible !

And I check vOptimizer.cc, in which shows that “// Send an error if FOM computation with list-mode data (it has no sense)”. So dose that means the list-mode data can not get the value of objective function?

Kind regards,

Bo Li

4

Yes I forgot to mention that.

For the moment, the likelihood and least-squares are computed and are valid only for histogrammed data.

I think that the computation of the listmode likelihood should be feasible considering equations as proposed originally by Barrett and colleagues. It would require some small modifications of the code thought.

Best
Simon