Information Technology Reference
In-Depth Information
Success in meeting the radiation amount prescribed to the tumor by a physician,
while meeting critical organ tolerances, requires modification of the intensities to be
delivered through each aperture. Due to the time consuming nature of modifying each
beamlet intensity value, optimization is inherently required for IMRT treatment plan-
ning. Current commercial treatment planning systems use simulated annealing to op-
timize each beamlet and can take anywhere from 30 minutes to 1 hour to meet the
prescription criteria.
Based on the results in brachytherapy as discussed above, the introduction of Har-
mony Search to the IMRT would provide a significant reduction in the time spent in
treatment planning. Additionally, a modification of the IMRT technique, known as
Volumetric Modulated Arc Therapy (VMAT), has been recently introduced in radiation
therapy departments across the globe, which modulates the intensities while the gantry
is rotating [16]. This would increase the possible decision variables (beamlet intensities)
anywhere from 5- to 10-fold as current IMRT treatments only use 5-9 beam angles,
while rotational therapy can be modeled as approximately 50-60 discrete beam angles.
Although the promise of VMAT would allow increased dose delivery to the tumor
and decreased dose to critical structures, the significant increase of computation time
to plan such a treatment is a perfect fit for Harmony Search to be applied to this new
technique in medical physics.
5 Conclusions
This chapter reviewed the novel application of Harmony Search as an optimization al-
gorithm to the field of medical physics. A DVH-based objective function was created
and used for the optimization simulation in HDR brachytherapy for prostate cancer.
Harmony Search and genetic algorithm were employed as optimization algorithms for
the simulation and were compared against each other for nine different patients. The
comparison between Harmony Search and genetic algorithm showed that Harmony
Search was over four times faster when compared over multiple data sets. The average
time per iteration was found to be faster for the genetic algorithm than for Harmony
Search due to the fact that the latter must randomly choose decision variable values
from the Harmony Memory to create a new solution vector once per iteration; the for-
mer chooses only two parents per iteration to create a new solution vector. Addition-
ally, using floating point values for dwell times, Harmony Search was still at least four
times faster than the genetic algorithm, corroborating the results for the integer mode.
However, in floating point mode, the simulation produces less than satisfactory results
and unless required, integer mode should be used. Since the GammaMed treatment unit
uses integer dwell times, integer mode is a perfect match. Finally, the optimal values
for the Harmony Search parameters (HMS, HMCR and PAR) were determined for the
HDR prostate brachytherapy simulation.
In conclusion, Harmony Search was shown to be quite capable as an optimization
algorithm for the use in HDR prostate brachytherapy. It is a suitable alternative to
existing algorithms. Coupled with the optimal parameters for the algorithm and a
multithreaded simulation, this combination has the capability to significantly de-
crease the time on time-intensive clinic problems such as brachytherapy, IMRT,
VMAT, TomoTherapy, and beam angle optimization.
Search WWH ::




Custom Search