Ponente
Descripción
Simulations of acquisitions in medical imaging techniques such as SPECT, PET, CT, Ultrasound, etc. is a fundamental tool in medical physics research. One of the important aspects of the simulation procedure is the use of the most realistic numerical models of the human body possible. In this regard, it is very helpful for these models to be segmentated in tissues and organs so that specific physical properties can be accurately inserted in the phantoms. The principal limitation for many applications is the availability of a large set of those detailed models.
One of the most commonly used numerical human phantoms, is the XCAT phantom simulator, developed by Duke university researchers. The anatomy in XCAT is defined using non-uniform rational b-splines (NURBS) and includes thousands of defined structures as well as parameterized models for the beating heart and respiratory motions. Its capacity to generate dynamic images allows the simulation of different medical images techniques that would not be possible with static models. In XCAT users can define numerous parameters to create normal and abnormal anatomical and motion variation, with specific properties and geometries. This flexibility makes it ideal for creating a large database of models for use in simulations. However the XCAT models are severely lacking in certain areas of the body, such as the breasts. Real breasts are composed of ducts, adipose tissue, lobules, etc. whereas the XCAT models do not show any internal structures.
On the other hand, the VICTRE package is a software developed as part of the Virtual Imaging Clinical Trials for Regulatory Evaluation (VICTRE) project of the Food and Drug Administration (FDA) in the USA. The VICTRE tools enable creating realistic breast phantoms and provide a complete simulated imaging chain for mammography and digital breast tomosynthesis, and it is an accessible tool. It allows modeling breast with different sizes and different proportion of fat and fibroglandular tissues.
In this work, we took advantage of the advanced image coregistration techniques provided by the SimpleITK python library to incorporate the VICTRE breast models into the full body XCAT phantom. We were able to generate female full body phantoms with realistic breast models (Fig. 1) which could be used for numerical simulations in the torso area. Some of these simulations include cardiac ultrasonography and positron emission mammography. The properties of the phantoms, and their capabilities will be described, as well as some metrics of the performance of the corregistration procedure.
