Calibration and Correction of Images

The calculated positions of the scintillation events are collected in a matrix, i.e. an image. When choosing the pixel size it is important to keep in mind that if the size of the pixel is too small, the statistics of the number of counts in the pixel will be weak. Typical matrix sizes: 64x64, 128x128, 256x256, etc. The raw images obtained after position determination are usually distorted – the extent and the characteristics of distortion depend on the applied position determination technique, but further corrections and calibrations are necessary in all cases.
The distortion can cause the image to be nonlinear (the image of a straight line is not a straight line) or inhomogeneous (homogeneous ‘lighting’ provides a structured image), edge effects may appear (blurring, compression), the image may not be calibrated in energy, or appear in the middle. The calibrations fixing these defects are:

• Spectrum (energy offset) calibration
• Setting of the PMT amplifiers (autotuning)
• Spatreg (Spatial image offset) calibration
• Linearity calibration
• Intrinsic (collimator-less) uniformity calibration

 
Spectrum calibration: the relationship between the channel number and the energy (keV) is determined for a given high voltage value using two isotopes (usually Tc-99m and I-131), assuming this relationship is linear.

Figure 1.

 
Autotuning: the adjustment of the amplification of the PMTs using a software so that the differences in the amplification of the individual PMTs are compensated. The course of the process: the camera is illuminated by a remote homogeneous source, then the spectrum of the events arriving in different regions above the PMTs are collected separately. As a result, spectra are recorded for each PMT, in which the ratio of the photopeaks can be determined. Afterwards, amplification factors, to which the amplifiers can be set, are calculated from these ratios in an iterative process.

Figure 2.: Uniformity image before autotuning

Figure 3.: Uniformity image after autotuning

 
The Spatreg calibration is necessary because the distortion and the size of the image typically changes if it is binned into a different energy window and it is not in the middle of the image matrix. First gain calibration is needed during calibration, which drags the image in the middle. Using a collimated source if the centre of the image is lit, the spot has to be in the geometrical centre of the matrix. If in case of an Anger camera this is not so, the amplification of the corner signals needs to be set. The second step is the spatial offset calibration using five gallium-67 sources. The gamma energies of the Ga-67 are 93 keV, 185 keV and 300 keV, on which three energy windows can be fitted. Our goal is to achieve that the positions imaged at every energy fall in the same place. In case of an Anger camera it can be accomplished by shifting the centre of gravity/energy. However, in this case the device is only calibrated to the measured energies and it is possible that at different energies it will position the sources in a different place. More accurate results can be achieved by applying linearity corrections at every energy.

Figure 4.: Spatreg test measurement using five Ga-67 sources

 
The aim of the linearity calibration is to ensure that the imaged distance the image of two point sources, the real distance of which from each other is d, will also be d regardless of their position in the field of view of the camera. To achieve this, linearity phantoms need to be used, which can be in the x or y direction, can consist of slits cut in a lead plate or a hole grid drilled in a lead plate. The image of the phantom is taken by placing it close to the camera and illuminating it with a remote point source (illuminating of course means irradiation by the source of radiation). The correction will be the vector field containing the radius vectors between the points of the distorted image and the expected linear image. During the subsequent image formation the calculated positions will be shifted by this position dependent vector, thus collecting the events into a matrix.

Figure 5.: The Anger image of a simulated linearity phantom in the y direction before linearity correction

Figure 6.: The Anger image of a simulated linearity phantom in the y direction after linearity correction

 
During energy calibration an energy correction map is taken based on local spectra. Due to the locally different scintillation light loss the photopeak (full-energy peak) appears at slightly different positions in the spectra drawn on the individual pixels. These deviations need to be corrected by shifting the energy of the incoming event after the position calculation, as a function of the position, based on the energy correction map. Then we apply energy filtering to this shifted energy. As a result of this calibration the photopeak becomes narrower.

The uniformity calibration corrects the inhomogeneity of the image. The collected images are corrected by the pattern (showing the structure of the PMT), as sensitivity, of the image taken of the homogeneous source.

Figure 7.: Homogeneous image before correction

Figure 8.: Homogeneous image after correction