few questions about castor

Dear Emanuele,

You can enable the verbose to see the progression of the sensitivity image generation (add the following option : -vb 2)

This computation performs a dual loop on all the detector elements to compute all possible lines of response, depending on the potential angle restrictions in the geometry file and datafile. So the duration will mostly depend on the total number of detectors and the image dimensions.

Once the sensitivity image is computed, it could be re-used for other reconstructions (provided these reconstructions use the same dimensions, projectors and image-based operations such as psf, etc…) using the option -sens sensitivity_file.hdr.

I will forward this answer to our user mailing-list as this is a question that other users could potentially have.


Hi Emanuele,

I have two comments in addition.

First, I strongly suggest to compile CASToR with multi-threading support (with CASTOR_OMP, see the documentation). Then you can use the option "-th 0" to tell CASToR to use as many cores as available in your computer, so a good improvement !

Second, beware of the gaussian filter parameterization ! When you set "gaussian,0.7,0.7,1.::psf", I suppose that you expect a gaussian with 0.7 FWHM in X, 0.7 in Y and 1 in Z. So then you should use "gaussian,0.7,1.,3.::psf", because the first parameter is for transaxial FWHM (so both for X and Y), the second parameter is for axial FWHM, and the third parameter is the number of sigmas of the gaussian filter that are kept into the integral of it in the convolution kernel. For the latter, we suggest 3 sigmas.


Dear Thibaud,

Many thanks. Our scanner has about 1.5M detectors, so it is normal that it takes all that time.



Dear Simon,

Thanks for the suggestions. Ok I misunderstood the option of the Gaussian smoothing.