Biomedical Engineering Reference
In-Depth Information
FIGURE 7.3
iSEG image segmentation tool and models that have been segmented based on MRI and CT images, respectively.
As the segmentation tool of HYCAT that has been developed
specifically for the purpose of hyperthermia treatment planning,
iSEG (see Figure 7.3) offers the possibility of flexibly combining
various segmentation techniques ranging from highly automatic
to highly interactive. The goal is that simpler tissues (e.g., bones
in CT) should be automatically segmentable while more complex
ones can be segmented with more user control. Segmentation
techniques include interactive watershed transformation, level set
methods, fuzzy connectedness, live-wire (competitive, hysteretic)
region growing, image foresting transformation, automatic thresh-
olding, brush drawing, and contouring [122, 127]. Topologically
flexible, adaptive inter- and extrapolation is available to automati-
cally suggest a segmentation based on previously segmented nearby
slices. Support for multimodal image data is provided. Dedicated
vessel segmentation is available for contrast enhanced CT. Noise
reduction and boundary enhancement filters are available for pre-
processing. Post-processing includes hole and gap closing routines,
surface and contour extraction and simplification, skin adding,
outline correction tools, connected component analysis, etc. The
conclusions of [122] are that different methods should be used
for different tissues: some tissues require interaction, but perhaps
not on every slice when interpolation is applicable; segmentation,
except for the simplest tissues, is mostly performed in 2D due to
speed concerns and the higher likelihood of volumes to include
regions where leakage can occur; hybrid methods combining con-
tour and area-based information behave more robustly; robustness
is also increased by using competitive region growing methods; and
a predefined scheme should be developed that specifies the order of
steps to be followed by a user to allow segmentation to be performed
on a routine basis by technical staff.
The choice of the segmentation technique does not solely
determine the quality and speed of the segmentation process.
It is equally important to pay proper attention to the type of
user interaction that is required and supported [133]. [67, 191,
193] discuss the impact of segmentation on treatment planning
results (e.g., accuracy, continuous properties vs. tissue specific
properties, and number of distinguished tissues.)
7.4 Electromagnetic Simulations
In the case of hyperthermia above a few 100 kHz, the spatial
distribution of the power dissipation in different tissues due to
ohmic losses respective to dielectric relaxation (molecular fric-
tion) can only be obtained by a full-wave solution of Maxwell's
equations for the entire system (i.e., environment, antenna,
water bolus, and patient). The dissipated power or the specific
absorption rate (SAR; see Figure 7.4) is calculated as
SAR = σ
E
2
/ρ
where σ is the electric conductivity,
E
the root-mean-square
(RMS) value of the total electric field, and ρ the tissue density.
FIGURE 7.4
(Left) Model of a patient in a head-and-neck applicator
and (right) the calculated SAR distribution for an unoptimized set of
antenna steering parameters (see Figure 7.5 for an optimized distribu-
tion). Notice the complex SAR distribution and the undesired SAR hot
spots in the vicinity of the spinal cord.
