266
Electromagnetic Fields in Biological Systems
5.2.1.2 Anatomically Based Models
In most numerical treatments, the entire human body or parts of it are modeled by cubic
cells (voxels). In each cubic cell of the mesh, the tissue properties are considered constant. By assigning the corresponding electric properties to each voxel, one can easily
model the anatomical tissues and organs.
In the majority of recent studies, the data for simulating parts of the body are taken
from magnetic resonance imaging (MRI) scans (Dimbylow and Gandhi 1991; Dimbylow
and Mann 1994; Gandhi, Lazzi, and Furse 1996; Gandhi and Chen 1995; Hombach et al.
1996; Lazzi and Gandhi 1997; Lu et al. 1996; Meier et al. 1997; Schoenborn, Burkhardt,
and Kuster 1998), though models based on computed tomography (CT) scans and anatomical images have also been encountered (Bernardi, Cavagnaro, and Pisa 1996). MRI
and CT provide gray-scale image data from the head to the feet of the human body as
several transverse slices at a designated spacing. The resolution in each slice is in the
order of several millimeters. MRI data are generally superior to CT data in identifying
interior tissues because of the high contrast images of the soft tissues. In order to be
used in numerical dosimetry, these digital data sets must be converted to the so-called
segmented version (Chen, Gelb, and Renaut 2003; Wells et al. 1996; Zankl and Wittman
2001). The translation of 3D data sets obtained by MRI or CT scans into a numerical
model is a complex and time-consuming activity that is difficult to perform with only
automatic procedures, such as contour recognition algorithms. This process inevitably requires intervention by experts in human anatomy, who are able to distinguish
both transitional and marginal regions. Even if software for automatic identification
is applied, manual verification or correction is required. It is worth noting that MRI
or CT produced in different laboratories inevitably contain differing discretizations.
Furthermore, the resolution of the medical imaging techniques is, presently, too high
for using their results directly in numerical modeling.
Early examples of anatomically based models include a torso model consisting of
16628 cells, each of side length 12.7 mm, used for SAR prediction (Sullivan, Borup, and
Gandhi 1987), as well as a model consisting of approximately 35000 cells, each of side
length 10 mm (as acquired from CT scans of a cancer patient), used for hyperthermia
treatment (Sullivan 1990). As computing power increases and computer resources get
less expensive, there is a trend to refine the numerical description of the space modeled
and move to more detailed anatomical structures. In contemporary models, the highest complexity used for modeling the whole human body is about 50 tissue types and
the finest resolution is about 1 mm. For example, in studies by Martinez-Burdalo et al.
(2009), a 3D high-resolution head mesh, developed from nuclear magnetic resonance
has been resized into a finer grid, without loss of anatomical details. The resulting model
has 1 × 1 × 1.25 mm 3 size cells and includes 18 different biological tissues.
Even though models with resolution on the order of 1 mm are becoming available,
models with 2–3 mm resolution are quite adequate for frequencies of 1–2 GHz (Gandhi
et al. 1992). However, special care is required to accurately model organs of particular
importance to the human health, such as the eyes, the parotid, the hypophysis gland,
etc. (Bernardi, Cavagnaro, and Pisa 1996; Okoniewski and Stuchly 1996). Better resolution is gained by refining the grid locally. However, higher resolution in human head
Electromagnetic Fields in Biological Systems
5.2.1.2 Anatomically Based Models
In most numerical treatments, the entire human body or parts of it are modeled by cubic
cells (voxels). In each cubic cell of the mesh, the tissue properties are considered constant. By assigning the corresponding electric properties to each voxel, one can easily
model the anatomical tissues and organs.
In the majority of recent studies, the data for simulating parts of the body are taken
from magnetic resonance imaging (MRI) scans (Dimbylow and Gandhi 1991; Dimbylow
and Mann 1994; Gandhi, Lazzi, and Furse 1996; Gandhi and Chen 1995; Hombach et al.
1996; Lazzi and Gandhi 1997; Lu et al. 1996; Meier et al. 1997; Schoenborn, Burkhardt,
and Kuster 1998), though models based on computed tomography (CT) scans and anatomical images have also been encountered (Bernardi, Cavagnaro, and Pisa 1996). MRI
and CT provide gray-scale image data from the head to the feet of the human body as
several transverse slices at a designated spacing. The resolution in each slice is in the
order of several millimeters. MRI data are generally superior to CT data in identifying
interior tissues because of the high contrast images of the soft tissues. In order to be
used in numerical dosimetry, these digital data sets must be converted to the so-called
segmented version (Chen, Gelb, and Renaut 2003; Wells et al. 1996; Zankl and Wittman
2001). The translation of 3D data sets obtained by MRI or CT scans into a numerical
model is a complex and time-consuming activity that is difficult to perform with only
automatic procedures, such as contour recognition algorithms. This process inevitably requires intervention by experts in human anatomy, who are able to distinguish
both transitional and marginal regions. Even if software for automatic identification
is applied, manual verification or correction is required. It is worth noting that MRI
or CT produced in different laboratories inevitably contain differing discretizations.
Furthermore, the resolution of the medical imaging techniques is, presently, too high
for using their results directly in numerical modeling.
Early examples of anatomically based models include a torso model consisting of
16628 cells, each of side length 12.7 mm, used for SAR prediction (Sullivan, Borup, and
Gandhi 1987), as well as a model consisting of approximately 35000 cells, each of side
length 10 mm (as acquired from CT scans of a cancer patient), used for hyperthermia
treatment (Sullivan 1990). As computing power increases and computer resources get
less expensive, there is a trend to refine the numerical description of the space modeled
and move to more detailed anatomical structures. In contemporary models, the highest complexity used for modeling the whole human body is about 50 tissue types and
the finest resolution is about 1 mm. For example, in studies by Martinez-Burdalo et al.
(2009), a 3D high-resolution head mesh, developed from nuclear magnetic resonance
has been resized into a finer grid, without loss of anatomical details. The resulting model
has 1 × 1 × 1.25 mm 3 size cells and includes 18 different biological tissues.
Even though models with resolution on the order of 1 mm are becoming available,
models with 2–3 mm resolution are quite adequate for frequencies of 1–2 GHz (Gandhi
et al. 1992). However, special care is required to accurately model organs of particular
importance to the human health, such as the eyes, the parotid, the hypophysis gland,
etc. (Bernardi, Cavagnaro, and Pisa 1996; Okoniewski and Stuchly 1996). Better resolution is gained by refining the grid locally. However, higher resolution in human head