A set of 4D pediatric XCAT reference phantoms for multimodality research

Abstract

Purpose:The authors previously developed an adult population of 4D extended cardiac-torso (XCAT) phantoms for multimodality imaging research. In this work, the authors develop a reference set of 4D pediatric XCAT phantoms consisting of male and female anatomies at ages of newborn, 1, 5, 10, and 15 years. These models will serve as the foundation from which the authors will create a vast population of pediatric phantoms for optimizing pediatric CT imaging protocols. Methods: Each phantom was based on a unique set of CT data from a normal patient obtained from the Duke University database. The datasets were selected to best match the reference values for height and weight for the different ages and genders according to ICRP Publication 89. The major organs and structures were segmented from the CT data and used to create an initial pediatric model defined using nonuniform rational B-spline surfaces. The CT data covered the entire torso and part of the head. To complete the body, the authors manually added on the top of the head and the arms and legs using scaled versions of the XCAT adult models or additional models created from cadaver data. A multichannel large deformation diffeomorphic metric mapping algorithm was then used to calculate the transform from a template XCAT phantom (male or female 50th percentile adult) to the target pediatric model. The transform was applied to the template XCAT to fill in any unsegmented structures within the target phantom and to implement the 4D cardiac and respiratory models in the new anatomy. The masses of the organs in each phantom were matched to the reference values given in ICRP Publication 89. The new reference models were checked for anatomical accuracy via visual inspection. Results: The authors created a set of ten pediatric reference phantoms that have the same level of detail and functionality as the original XCAT phantom adults. Each consists of thousands of anatomical structures and includes parameterized models for the cardiac and respiratory motions. Based on patient data, the phantoms capture the anatomic variations of childhood, such as the development of bone in the skull, pelvis, and long bones, and the growth of the vertebrae and organs. The phantoms can be combined with existing simulation packages to generate realistic pediatric imaging data from different modalities. Conclusions: The development of patient-derived pediatric computational phantoms is useful in providing variable anatomies for simulation. Future work will expand this ten-phantom base to a host of pediatric phantoms representative of the public at large. This can provide a means to evaluate and improve pediatric imaging devices and to optimize CT protocols in terms of image quality and radiation dose.

A personalized, Monte Carlo‐based method for internal dosimetric evaluation of radiopharmaceuticals in children

“…Later, Norris et al. developed a set of four‐dimensional (4D) realistic pediatric models based on the hybrid extended cardiac‐torso (XCAT) phantom and the use of nonuniform rational B‐splines (NURBS) that incorporate organ motion, such as respiratory and cardiac motion . More recently, Xie et al.…”

Two Realistic Beagle Models for Dose Assessment

“…Realistic body models have been developed by several groups (e.g., Petoussi-Henss et al 2002;Lee et al 2005;Stabin et al 2012;Segars et al 2013;Norris et al 2014). The latter series used a modeling technique developed by Segars (2001), defining the organs and structures with NURBS surfaces.…”

Pediatric personalized CT-dosimetry Monte Carlo simulations, using computational phantoms

“…A DCT-10 ionization chamber was modeled in GATE in order to validate the simulated X-ray source. Furthermore, clinical pediatric CT data were processed (segmented) to be imported in GATE and the XCAT [6] and IT'IS [7] anthropomorphic phantoms were used as reference models for the dosimetric simulations. In order to have accurate and realistic simulations the need of high number of simulated particles (a clinical study is in the order of ~10 13 particles) as well as fine resolution on the voxelized model ≤ 2.0 mm is apparent.…”