Back
to mini-symposia index
12)
Minimizing the adverse effects of uncertainities in
external radiation therapy
Principal organiser: Dr. Yuriy Zinchenko
Advanced Optimization Laboratory, McMaster University &
Department of Radiation Oncology, Princess Margaret Hospital,
CAS, McMaster University &
Department of Mathematics, University of Calgary
Summary:
Motion-induced and structural uncertainties in radiation therapy
severely limit the efficacy of the treatment. We target minimizing
adverse effects of uncertainties in IMRT during the treatment
planning using a combination of image guidance and novel large-scale
robust optimization techniques, pioneered by our group 5 years
ago.
Robust optimization allows exploiting not only spatial, but
also temporal characteristics of the uncertainties, in order
to produce more aggressive treatment plans. Computational studies
indicate superiority of this approach over the traditional treatment
planning, however, further investigation is required.
The minisymposium is intended for radiation physicists, oncologists,
treatment planning software developers and applied mathematicians.
Speakers:
1. Laura Dawson, Department of Radiation Oncology, Princess
Margaret Hospital
High-precision radiotherapy for hepatobiliary cancer:
the effects of uncertainties.
The
projected increase in the incidence and mortality of thoracic
and abdominal cancers is an important health concern for
Canadians. For many diseases, radiation therapy is a proven
treatment modality, and nearly half of all cancer patients
receive radiation therapy, either as the sole treatment
modality or in combination with surgery and/or systemic
pharmaceutical agents, e.g. chemotherapy. Important technological
advances, such as intensity modulated radiation therapy
(IMRT), have solidified or expanded the role of radiation
therapy in improving control of cancers and reducing toxicity,
notably for head and neck cancers and prostate cancer. For
disease in the thorax and abdomen, however, the effectiveness
of radiotherapy has been limited by the sensitivity of non-tumorous
lung and liver tissues, and confounded by the technical
challenges associated with breathing-related organ movement
and patient setup uncertainty. Highly tailored treatments,
where the high dose volume matches the shape and position
of the cancer, minimize the dose to healthy tissue but also
create the risk of a geometric miss of the disease. Without
effective targeting there is a risk of damaging healthy
tissues indiscriminately and failing to control the disease,
even using IMRT technologies. We believe there is a compelling
need to develop properly targeted IMRT in order to yield
significant gains in the control of lung and liver cancer,
including cancer originating in the lung and liver, as well
as metastases to these sites from colorectal cancer and
breast cancer.
We discuss how the uncertainties impact the efficacy of
the radiation treatment for primary and secondary liver
cancers – a particularly challenging site for this
type of treatment modality.
2.
Tamás Terlaky, School of Computational Engineering
and Science, McMaster University
Large-scale convex and robust optimization: introduction
and applications
Two
decades of Interior Point Method research opened new avenues
to solve large scale optimization problems. Novel modeling
methodologies, such as convex conic and robust optimization
were developed. Now we are able to model and routinely solve
large-scale problems with high practical importance. In
this talk we review the state-of-art of large scale convex
optimization, discuss the Ben Tal - Nemirovski robust optimization
model, review some novel applications and illustrate the
power of the state-of-the art SeDuMi conic linear optimization
software package.
3.
Kristy Brock, Department of Radiation Oncology, Princess
Margaret Hospital
Novel deformable image registration techniques to facilitate
classification, targeting, and monitoring of tumor and normal
tissue response using finite element modeling.
As
more pretreatment imaging becomes integrated into the treatment
planning process and full three-dimensional image-guidance
becomes part of the treatment delivery the need for a deformable
image registration technique becomes more apparent. Several
novel finite element model-based multi-organ deformable
image registration methods, e.g., MORFEUS, have been developed.
A common basis of these methods is twofold: first, individual
organ deformation can be accurately modeled by deforming
the surface of the organ at one instance into the surface
of the organ at another instance and assigning the material
properties that allow the internal structures to be accurately
deformed into the secondary position and second, multi-organ
deformable alignment can be achieved by explicitly defining
the deformation of a subset of organs and assigning surface
interfaces between organs. The feasibility and accuracy
of the method was tested in a clinical setting. We discuss
the potential applicability of these models.
4.
Yuriy Zinchenko, Department of Mathematics, University
of Calgary
Convex and robust optimization for IMRT treatment planning
Our
goal is to develop treatment planning strategies that combine
the assessment and intervention capabilities of IGRT with
novel robust and adaptive large-scale optimization techniques
to maximize the use of available patient information while
limiting the need for real-time corrective intervention. The
process of IMRT treatment planning, besides acquiring the
initial medical imaging data and setting proper clinical objectives,
typically involves formulating and solving an inverse problem
that models treatment delivery. Optimization techniques are
used commonly for the inverse problem, in order to attain
the treatment delivery that is the best possible fit to specified
clinical objectives.
To reduce the computational burden associated with IMRT treatment
planning, we propose a convex optimization model that incorporates
the desired dose distribution requirements. To minimize the
negative effects of the uncertainties, we propose a convex
robust model, which is partially based on probabilistic reading
of the constraints. Both models fall within the framework
of structured convex optimization, and, thus, are potentially
amenable to Interior Point Methods. We highlight current challenges
associated with both of the approaches
|
|
