Biomedical Engineering Reference
In-Depth Information
interspersed within normal tissue, or may be supported on normal
tissue stroma, and the PTV certainly includes normal tissues outside
the CTV. To avoid damage to these normal tissues, one often places
an upper bound on the dose within the PTV. The second factor is the
dose inhomogeneity. Even if one is pretty confident in the model
used to estimate TCP or EUD, one may not trust it enough to allow it
to accept a highly inhomogeneous dose distribution within the target
volume. Thus, it is common to place a dose constraint on the dose
inhomogeneity. This constraint can be achieved by placing upper and
lower bounds on the dose within the target volume, or by placing a
constraint on the difference between the EUD and, say, D
mean
.
Estimating normal tissue response
As discussed in Chapter 6, in manual plan evaluation one tends to
separately inspect each organ at risk (OAR) and the remaining
volume at risk (RVR) to assess the impact of the dose distribution
delivered to each of those volumes of interest. The same approach is
fully appropriate to, and widely used in, computer-driven planning.
For each OAR, one computes a quantity such as V
20Gy
, or EUD, or
D
max
, or NTCP, and assigns a sub-score to that OAR based on the
computed value. The sub-score is then either used in computing the
value of the score function, or is used in connection with a dose or
dose
volume constraint.
In practice, it is often the case that the normal tissue constraints do not
adequately force the optimization process to produce acceptable
results. For example, the dose may not fall off as rapidly as desirable
outside the target volume because the identified normal tissue
constraints are not strong enough, or are not defined for all tissues
outside the target volume. De Neve
et al
. (2006) have given a good
discussion of many of the tricks that can be used to achieve
satisfactory results. These include defining a shell or shells around
the target volume within which additional constraints can be applied,
and defining a “virtual” normal tissue in a region in which a hot spot
occurs which will drive the dose down in that region when the
optimization process is repeated.
−
Combining tumor and normal tissue responses
In constrained optimization with hard constraints, one does not have
to combine different elements; one optimizes one element, and places
constraints on all others of interest. Score optimization is different;
