Could you help us answer a question regarding the reconstructed image units in CASToR?
Basically, a student of mine reconstructed a Monte-Carlo simulation of a NEMA IQ phantom which should exhibit activities from 5.3kBq/ml in the background to around 21.2kBq/ml in the hot spheres. The reconstruction is performed using CASToR with a voxel size of 2x2x2mm3.
When visualizing the reconstructed images in amide we obtain values in the hot spheres in the order of magnitude of 4e-4, but we do not know which unit the images have and whether they are dependent on the voxel size. The 4e-4 should match the input of 21.2kBq/ml if we did not make a mistake using either GATE or CASToR.
Therefore, how can we convert the values of the reconstructed images into values of activity per ml (kBq/ml)?
While we are asking about units, what is the unit of the sensitivity map? The sensitivity map has values of about 1e5-1e6.
Thanks for your help and best regards,
Jan
P.S.: Seems like my student cannot post to the mailing list, so I tried on his behalf. If you need the exact command line arguments, I will ask my student and deliver them subsequently.
I read your question and I have the same curiosity, I need to know the units of the reconstructed images in castor to follow up on my thesis project, I see that there was no response a year ago. Would you be so kind to tell me what you found out about the subject?
Hi,
By default, the units in the reconstructed images are basically count/voxel.
The calibration factor (defined in the histogram) is to normalize the sensitivity image values in order to get the resulting images in kBq/mL or any meaningful value. One potential way to compute such factor is to use a uniform object such as a cylinder with known uniform activity concentration. From the estimated value in the reconstructed images and the known activity concentration, one could derive a calibration factor.
The sensitivity image values are computed by projecting all the line of response of a normalization or histogram datafile (if provided) or all the available lines of response from the geometry (default). Due to the number of lines, the sensitivity image values tend to be rather high for usual systems (>1e5), resulting in very low reconstructed image values if the calibration factor is not fixed.
Firstly, is calibration inversely proportional to time and voxel spacing？So it would be better if I could compute the calibration factor per second, then I could get the corresponding calibration factor by dividing with the scanning time.
Secondly, should I make a mask to fit the cylinder? Cuz, there are noises in full image(and I found add min angle and small FOV causes the highlight at the edge of FOV, as shown in the following figure， raw image 280x280x280, m means min angle, f mean 200x200x280). And should I do all correction before computing calibration?(normalization and attenuation are necessary)
Are you intending to estimate the calibration factor to get images in Becquerel per mL ?
First, in CASToR, the intensity of the reconstructed image is natively divided by the scan duration (if scan duration is specified in the datafile header), to get counts per second. As a consequence, the calibration factor is independent of the scan duration. Same for decay : the intensity of the reconstructed image is natively corrected for decay and radionuclide b+ branching ratio (if the radionuclide is specified in the datafile header). As a consequence, the calibration factor is independent of the radionuclide.
Second, to get Bq/mL, you have to apply all corrections (normalization, random and scattered coincidences, attenuation, dead-time) during the reconstruction. Then, you will get a calibration that is independent of the source.
Yes, I am intending to get images in Bq/mL! Thanks for your reply! It describes very clearly!
So I add scan time and the radionuclide to the datafile header, then add normalization and mumap datafile to recon from GATE root(only true coincidence) during the reconstruction. I will try to compute the calibration factor as activity concentration devided by total voxel value as soon.
