Skip to content

Using Inveon CT data

Jump to bottom
villekf edited this page Apr 20, 2021 · 2 revisions

This page describes the contents of the open Siemens Inveon CT data and how to use it. The data can be obtained from https://doi.org/10.5281/zenodo.4646835.

Files included

NEMA_S6_18F_Inveon_190M_counts_XAT_Attenuation-CT_NEMA_s6_v1.cat is the raw projection data. Contains first a header section, then the dark measurements, then the light measurements and finally the actual projection data. NEMA_S6_18F_Inveon_190M_counts_XAT_Attenuation-CT_NEMA_s6_v1.cat.hdr is the corresponding header file that contains the necessary scanner information and can be input to loadInveonCTData when prompted.

NEMA_S6_18F_Inveon_190M_counts_XAT_Attenuation-CT_NEMA_s6_v1.ct.img is the FDK reconstructed image of the projection data. Reconstructed by the Siemens Inveon Acquisition Workplace software. The header file NEMA_S6_18F_Inveon_190M_counts_XAT_Attenuation-CT_NEMA_s6_v1.ct.img.hdr contains the reconstruction parameters.

Inveon data in OMEGA

This data can be used without modifications with Inveon_CT_main.m. Running this file prompt you to select the projection header file (NEMA_S6_18F_Inveon_190M_counts_XAT_Attenuation-CT_NEMA_s6_v1.cat.hdr). Note that, by default, the script will first run the built-in reconstructions and then the forward/backward class examples. It is recommended to comment all, but one of these sections (they are separated by %%). The data has already been binned 4x4 and, as such, selecting higher binning value will bin the data even more.

The phantom data is a step-and-shoot cone beam measurement that contains three bed positions. This data was originally used for PET attenuation correction. You can select any or all bed positions. Axial FOV and axial number of pixels will be automatically adjusted.

If you use other Inveon CT data, simply modify the necessary variables accordingly. All scanner parameters will be automatically loaded from the header file.

For reconstruction, it is recommended to use random subset sampling (subset_type = 3) if you use subsets. Furthermore, reconstructing all three bed positions can take hours on the CPU, depending on the number of iterations and core count.

Clone this wiki locally