XCAT phantom data

DB Dominik F. Bauer
TR Tom Russ
BW Barbara I. Waldkirch
CT Christian Tönnes
WS William P. Segars
LS Lothar R. Schad
FZ Frank G. Zöllner
AG Alena-Kathrin Golla
ask Ask a question
Favorite

The XCAT model provides highly detailed whole-body anatomies. Organ masks can be easily obtained within the XCAT framework. Since CycleGANs maintain the geometry provided by the XCAT, the organ masks can be used as segmentation masks in the synthesized images. The phantom includes female and male models for varying ages. The heart beat and respiratory motions can be simulated and displacement fields of these motions can be generated. The anatomy and motion can be adapted by various parameters. This allows the creation of highly individual patient geometries. For the XCAT training data we generate one XCAT volume per XCAT model for each modality with 56 different models of varying ages. The XCATs include the whole liver and are generated with the same voxel spacing, windowing and normalization as the resampled patient data. Arms are included only in the MRI XCATs.

The XCAT phantom provides attenuation coefficients for all organs. We vary the simulated tube energy of the CBCT and CT phantoms from 90–120 keV in steps of 5 keV. This leads to a variation of attenuation coefficients in the phantoms. Afterward, those are transformed into Hounsfield Units. To obtain CBCT and MRI XCAT data, we need to convert the CT XCAT. For the CBCT XCAT, we apply a field of view mask obtained from the patient CBCTs, which is centered on the liver. For the MRI phantoms, we replace the attenuation coefficients for each organ with simulated MRI values using the signal equation for the VIBE sequence. It ensures that the MRI signal is initialized with realistic values matching the MRI training data. This enables us to use the aforementioned intensity and gradient loss for the generation of synthetic MRI images, since the transformation with the CycleGAN is now monomodal. The signal intensities (SI) for the VIBE sequence in terms of acquisition parameters repetition time TR, echo time TE, and flip angle α and tissue-specific T1, T2 relaxation times, and proton density ρ is given by:

We calculate the MRI intensity for all 44 abdominal organs present in the XCAT. The imaging parameters TE=4.54 ms, TR=7.25 ms, and α=10 are obtained from the patient VIBE scans. The values for the proton density ρ are taken from [23]. T1 and T2 relaxation times for 3 T for blood and the spinal cord are obtained from [18] and the rest from [6]. For organs with no available T1, T2 or ρ, we use values of similar organs. To simulate some organ variability, we randomly vary T1, T2, and ρ by ±5% using a uniform distribution.

Do you have any questions about this protocol?

Post your question to gather feedback from the community. We will also invite the authors of this article to respond.

post Post a Question
0 Q&A