Using Inveon PET data

Table of Contents

• Files included
• Inveon data in OMEGA
  • Remarks on reconstruction

This page describes the contents of the open Siemens Inveon PET data. The data can be obtained from:

Files included

FDG_Naive_R01_PET_Em_F18_60min_v1.lst is the raw list-mode data. Each event is stored in binary format with the tag related information. For decrypting the list-mode data itself, you should contact Siemens. A header file is also included that has some general information on the measurement and the device.

FDG_Naive_R01_PET_Em_F18_60min_v1.scn contains the sinogram created by the PET machine/acquisition computer itself. Randoms correction has already been included in the form of subtraction. This sinogram is dynamic and contains 28 frames. A header file is included that contains more detailed information on the file.

FDG_Naive_R01_PET_Em_F18_60min_v1_sct.scn this is the machine created dynamic SCATTER sinogram. Format is exactly the same as the regular sinogram. This should be used for potential scatter correction.

FDG_Naive_R01_PTX_At_10min_v1.atn this contains the attenuation data. This cannot be used as-is. In OMEGA this measurement data is reconstructed and then used in the attenuation correction, by using the reconstructed images as the attenuation images.

Norm_PNM_Normalization_with_Cylinder_v1.nrm is the normalization data. The inverse of this should be multiplied with the sinogram for normalization correction (when normalization is not included in the reconstruction). In OMEGA, this can be used as the user normalization.

Inveon data in OMEGA

When using the Inveon data in OMEGA, one should use Inveon_PET_main.m file. All the geometry and sinogram values have been adjusted just as the defaults are. This file includes support for GATE data, so when using the list-mode or sinogram data all the GATE specific sections can be ignored.

Inveon data format is chosen in the second block/section below MACHINE PROPERTIES (INVEON DATA). Here you can specify whether to use GATE data (options.use_machine = 0), the list-mode (.lst) data (options.use_machine = 1, default) or sinogram data (options.use_machine = 2). For the list-mode case, the user will be prompted for the location of the list-mode data file when load_data is run. For the sinogram case, the prompt will be during sinogram formation (form_sinograms). Using GATE data works just as with any other GATE data.

When using any corrections, same rules and features apply as with any data.

If options.randoms_correction is set to true, then the delayed coincidences will be automatically loaded when the prompts are loaded as well.

When using scatter correction, you can use the provided scn-files when prompted for the scatter data. Scatter data will be prompted FIRST, if you have it enabled.

In static cases, the dynamic sinograms are summed together. However, if dynamic reconstruction are performed, then the number of time steps need to be same as in the sinograms (28) or less. Raw list-mode data is NOT affected by this; it can have any number of time steps.

Normalization is included if options.normalization_correction = true and options.use_user_normalization = true. The user will then be prompted for the normalization file (.nrm) either during sinogram creation (if options.corrections_during_reconstruction = false) or after image reconstruction is initiated.

For attenuation correction, first options.attenuation_correction must be set to true and secondly the section containing the function (or the function itself) attenuation_correction_factors needs to be run. The user will then be prompted to select the provided .atn-file. The function will then save a mat-file containing the attenuation images reconstructed from the attenuation sinogram. The output is the path to this mat-file which is needed during the image reconstruction phase. By default, no additional actions are needed from the user. For future use, it is recommended to modify options.attenuation_datafile path after options.attenuation_correction to the path of the saved attenuation datafile. Filename itself is enough if the file is in MATLAB search path.

Remarks on reconstruction

The recommended (and default) resolution for the final image is 128x128x159. Arc correction is recommended as it decreases aliasing artifacts. Arc correction uses parallel computing toolbox (parfor) if it is available.

Using higher resolution (256x256x159) causes aliasing artifacts with all projectors though orthogonal distance-based ray tracer in 3D mode only has some artifacts on higher iterations. Using other projectors is possible if the sampling rate (options.sampling) is increased. Using a sampling rate of 4x should remove aliasing artifacts even with 1 ray Siddon. Arc correction may not be required when using increased sampling rate.

Another solution to prevent aliasing artifacts should be the use of regularization, but this is untested.

It is not recommended to use 3D orthogonal projector on CPU with 256x256x159 resolution.