The XCAT phantom is a digital hybrid that uses B-Splines fitted to cadaver CT data from The Visible Human Project to define the anatomy of the human torso. Based on the modelled anatomy, the phantom can simulate several imaging modalities including fluoroscopy, CT and PET. It allows the selection of parameters for processes such as image acquisition and image reconstruction. This means that new imaging techniques can be tested and evaluated virtually, reducing the need for imaging studies that use patients or physical phantoms.
A more realistic image
The XCAT phantom models respiratory motion, but to date has only been able to simulate regular breathing patterns. Irregular breathing patterns, where the duration and amplitude of the respiratory cycle may vary significantly from cycle to cycle, provides a more clinically relevant test for image processing algorithms.
"When using a digital phantom to investigate the effects of motion on imaging techniques, it is critical that the phantom has as realistic a breathing motion as possible. Using too simple a respiratory motion model will not accurately reflect an algorithm's performance in a realistic clinical situation," explained senior author, physicist John Lewis, of the Brigham and Women's Hospital, Dana-Farber Cancer Institute and Harvard Medical School (Boston, MA).
"The phantom can be used to test and improve the ways in which irregular patient motion is accounted for in CT reconstruction and dose calculations. For example, in CT reconstruction it can be used to compare various 4DCT sorting algorithms or acquisition protocols," said Lewis.
In the previous version of XCAT, tumour motion was controlled indirectly through the regular respiratory motion of the diaphragm and chest surface. The researchers adapted the phantom so that any individual 3D tumour motion could be input directly into the phantom. However, the increased flexibility came with a cost: the modification uncoupled tumour motion from the chest and diaphragm motion. To correct for this, the researchers developed an algorithm that synchronizes the tumour motion with the two structures over the respiratory cycle.
In vivo tumour motion data were obtained for input into the phantom using X-ray fluoroscopy images of 10 patients with lung tumours. The researchers were able to use the measured data to test the efficacy of the synchronizing algorithm by simulating a 4D CT image set and comparing tumour motion with diaphragm motion before and after synchronization. The algorithm performed well: the normalized tumour displacements matched the diaphragm displacements.
Synchronization success
"Verification of the phantom's adaptive motion synchronization capability confirms its suitability when recorded independent tumour motion is available," said physicist Pankaj Mishra, first author on the study.
The researchers assessed the accuracy of the modified phantom by comparing the measured in vivo tumour motion data with that observed in a 4D CT image set simulated by the phantom. Good agreement between the measured and simulated motion was observed, with average root mean square errors of 0.29 ± 0.04 mm (x direction), 0.54 ± 0.17 mm (y direction) and 0.39 ± 0.06 mm (z direction).
To demonstrate the phantom's utility for algorithm assessment, the researchers also used image data simulated by the phantom to test the performance of a particular 4D CT sorting algorithm. The phase sorting technique uses internal anatomical features in the simulated images to allocate each to a phase in the respiratory cycle and was successfully applied to the simulated image data.
The group members were pleased with the modified phantom's overall performance. "The two applications carried out in this work [of 4D CT simulation and phase sorting] demonstrate the possible usage of modified XCAT phantom for simulating clinical experiments using the phantoms," said Mishra.
The MATLAB(C) code used to adapt the phantom is available from the authors on request.
• Related articles in PMB
Adaptation and applications of a realistic digital phantom based on patient lung tumor trajectories
Pankaj Mishra et al Phys. Med. Biol. 57 3597
On a PCA-based lung motion model
Ruijiang Li et al Phys. Med. Biol. 56 6009
Development of a model of the coronary arterial tree for the 4D XCAT phantom
George S K Fung et al Phys. Med. Biol. 56 5651
4D CT sorting based on patient internal anatomy
Ruijiang Li et al Phys. Med. Biol. 54 4821