The Virtual Family: A set of anatomically correct whole-body computational models | Center for Devices and Radiological Health

Catalog of Regulatory Science Tools to Help Assess New Medical Devices

This regulatory science tool presents a computer model of four highly detailed, anatomically correct whole-body models of an adult male, an adult female, and two children.

*VF1.0:

*VF2.0: https://itis.swiss/virtual-population/virtual-population/vip2/External Link Disclaimer

Technical Description

The Virtual Family (VF) is a set of four models that are based on high-resolution magnetic resonance imaging (MRI) data of healthy volunteers [1]. Organs and tissues of the VF 2.0 models are represented by three-dimensional, highly detailed computer-aided design (CAD) objects without self-intersections and gaps. The CAD objects allow the models to be meshed at arbitrary resolutions without loss of small features.

The following two VF model versions are available:

VF 1.0: The VF 1.0 models include segmentation of approximately 80 high-resolution organs and tissues. The Virtual Family Tool can be used to discretize and export the CAD objects in a generic voxel-based format. All four VF 1.0 models and the Virtual Family Tool are provided free of charge (except for handling fees) to the scientific community for academic purposes only.
VF 2.0: The VF 2.0 models consist of simplified CAD files optimized for finite-element modeling in any third-party platform. These models are based on a new, high-end generation of VF models that have been re-segmented at finer resolution to afford a higher degree of precision and anatomical refinement, as well as improved structural continuity of approximately 300 organs and tissues. For the purpose of simplification, these structures are combined into 22 high-resolution tissues. The VF 2.0 models are available free of charge (except for handling fees) to everyone.

*Virtual Family models (Age, Sex, Height, Weight, BMI)

DUKE: 34-year-old male, 1.77m, 70.2 kg, 22.4 kg/m2
ELLA: 26-year-old female, 1.63m, 57.3 kg, 21.6 kg/m2
BILLIE: 11-year-old female, 1.49m, 34.0 kg, 15.3 kg/m2
THELONIOUS: 6-year-old male, 1.16m, 18.6 kg, 13.8 kg/m2

Intended Purpose

Currently, the VF models are used for electromagnetic, thermal, acoustic, and computational fluid dynamics (CFD) simulations. Examples of applications of electromagnetic and thermal simulations are the assessment of the safety of active and passive medical implants in an MRI environment and the evaluation of the safety and efficacy of ablation devices [2]. Electromagnetic and thermal simulations have been performed on the entire set of VF models and additional models of children to calculate the whole-body averaged and local specific absorption rate (SAR) during exposure to 1.5 and 3T whole-body MRI coils [3,4]. These electromagnetic and thermal simulations have also allowed the evaluation of the safety of multi-channel transmit radio frequency MRI coils [5]. An example of application of electromagnetic and CFD simulations is the assessment of the applicability of the magneto-hemodynamic effect as a biomarker for cardiac output [6]. Acoustic simulations have been performed to assess the impact of the human anatomy on the focus location, shape, and intensity of ultrasound waves during focused ultrasound treatment [7].

Testing

VF1.0 is the first generation of models, originally developed for electromagnetic exposure evaluations. About 80 different tissues and organs were identified and segmented for each model [1]. The segmentation was carried out organ by organ or tissue by tissue. Although the image processing software significantly facilitated the segmentation process, the correct identification of tissues required 3D visualization of the organ shape. Several iterations per organ were necessary to achieve a satisfactory result. The angiographic image sets of the vessel trees supported the segmentation of the blood vessels of the two adult models. To test the segmentation quality, International Commission on Radiological Protection (ICRP) reference phantoms’ organ masses were used. There is good agreement between the organ masses of the ICRP reference phantoms and the obtained models for most tissues and organs. Certain deviations can be attributed to the segmentation uncertainties—mainly of the digestive tract. Deviations of the masses of those tissues, which show up clearly in the MR images, are due to individual properties of the volunteers [1].

Limitations

Despite the high contrast and resolution of the MR images, certain compromises during the segmentation were inevitable. Although the outlines of most of the tissues in VF models could be constructed with an accuracy of 1 to 2 pixels of the original images, the limitations of the models are as follows.

Blood vessels with a diameter of less than 2 mm were not included [1] (Recently the Virtual Population 3.0 became available [8]. Blood vessels with a diameter of less than 1 mm is available at the Virtual Population 3.0).
Nerve tissues only include the spinal cord and the optic nerve [1] (the Virtual Population 3.0 includes initial sections of the spinal nerves (about 2 cm) as well as the sciatic and tibial nerves [8])
Despite the administration of the anticholinergic agent to the adult volunteers, the imaging quality of the digestive tract was not sufficient for a complete segmentation. The intestinal ducts and the lumina could not be reconstructed completely; the pancreas could only be reconstructed for the male adult model [1].
The cortex of the kidneys and the walls of the stomach and the gall bladder were indistinguishable from the medulla and the lumen, respectively. They were segmented using a fixed thickness of five pixels of the MR images [1].
The salivary glands have not been segmented separately [1].
Only small fractions of the diaphragm were visible in the images. It has therefore been reconstructed with the help of anatomical atlases [1].
The bone marrow of the two female models has not been segmented for the entire body [8]. Since white and yellow marrow cannot be told apart in the MR images, the bone marrow has been segmented as yellow marrow in the male models.
The periosteum has not been segmented separately for all models [8].
The extrathoracic region has not been identified as a separate region [1].

The VF2.0 models are distributed as surface meshes in stereolithography (STL) format. Each tissue is stored as a separate surface mesh, which is guaranteed to be closed, but may contain non-manifold edges or vertices. Note also that some software tools are unfortunately not able to deal with non-manifold surfaces.