|
WORKSHOP AGENDA: February 7 - February 9, 2002
Feb 7: Evening Session: "Primer: Background Material"
6:30pm -- Refreshment and ligh hors D'oeuvres
7:30pm -- Welcoming remarks (NCI/NSF)
7:45 -- Primer talks:
- Joe Deasy Ph.D. Physics : - Radiotherapy Physics(40 min)
Really simplified radiation physics to understand radiation
therapy treatment planning (photon beams & the exponential law of attenuation).
Plus introductory comments about intensity modulated radiation therapy;
beams vs. beamlets; cell kill; tumor control; and normal tissue complications.
- Mark Langer M.D. Clinical: - Radiation Oncology (30 min)
How a physician reads a treatment plan.
What one looks for in the dose distribution.
Acceptance criteria, and guidelines for comparisons.
The effect of uncertainty on plan comparisons.
When will a plan be recognized as improved.
- Ron Rardin Ph.D. OR: - Operations Research/Optimization (50 min)
How an optimizer thinks about planning problems. Parameters,
decision variables, constraints and objectives. Linear vs.
nonlinear, continuous vs. integer, deterministic vs. stochastic,
and single vs. multi-objective models, and their implications for
tractability. Local and global optimal solutions. Relaxations and
bounds. Penalty methods.
Feb 8: Morning Session: "Radiation Treatment: Objectives, Formulations
and Clinical Implications"
7:30am -- Continental breakfast
8:00am -- 3 talks (each 20 min + 10 min Q&A)
-
Mike Zelefsky, M.D. The objectives for prescribing a
treatment plan.
Specification of constraints and preferences.
Use of "Constraint Templates".
Choice of Importance Factors - does it matter to the
physician?
-
Tom Bortfeld, Ph.D. Physicist. Present IMRT optimization
algorithms: principles, potential, and limitations.
Topics include inverse approach vs. forward calculation;
objective functions: "physical" and "biological";
optimization techniques: gradient, annealing,
maximum likelihood;
aperture optimization;
optimization and organ motion;
Multi-criteria optimization.
- Larry Marks, M.D. How the aims of a physician are conveyed
to treatment planners and the use of non linear response
functions. Dose, dose volume and nonlinear response
functions as predictors of treatment effect. Are they
reliable? How should they be used? What level of accuracy is
needed for their clinical acceptance and how can it be
obtained?
9:30 -- Break (15 minutes)
9:45 -- 3 talks (each 20 min + 10 min Q&A)
- Eva Lee, Ph.D. OR. Optimization with multiple objectives. Major points include:
importance of objectives/criteria; ranking, weights and weighted sum on
objectives; multi-attribute decision making via goal programming; multilevel
programming; and the penalty method.
- Avi Eisbruch M.D. Choosing among alternatives and deciding
on a treatment prescription. What are the physician's goals?
How does the physician balance among competing goals?
Development of prescription requirements for multi-
institutional protocols to meet different preferences of
physicians
- Dick Fraass, Ph.D. Physicist. Differences
between plan evaluation and the optimization
problem statement, and the difference it makes.
Discussion includes criteria of plan optimality; clinical criteria vs.
objective functions and constraints; comparing 'optimized' treatment
plans.
11:15 -- discussion panel (30 minutes)
Lunch and Keynote Presentations
12:00 -- 2 talks (each 30 min + 10 min Q&A)
- Ellis Johnson, Ph.D. OR. Optimization in Airline Scheduling,
Challenges and Successes:
How integer programming techniques are used in airline industry.
Several successful stories: crew scheduling, recovery and yield
management.
Numerical challenges for optimizers.
- Cliff Ling, Ph.D. Physicist. The Future of Radiation Oncology
and Cancer Research. New sources of information and images.
Afternoon Session: "Some Proposed Strategies: Strengths and Limitations"
2:00pm -- 3 talks (each 20 min + 10 min Q&A)
- Steve Billups, Ph.D. OR. Minimum-Support Solutions for
Radiotherapy Planning
Computer-generated plans for radiation treatment sometimes
involve an unnecessarily large number of gantry angles. Such
plans are time consuming and expensive to administer because
each distinct angle causes a delay while the gantry is
repositioned. To address this issue, we consider an
optimization model that generates minimum-support solutions
to the treatment planning problem--that is, solutions
involving a minimum number of gantry angles.
- Joe Deasy, Ph.D. Physicist. The IMRT optimization problem
statement.
How can the optimized treatment planning problem
statement be made more relevant to clinical goals? How can
inverse planning engines be made to be more steerable or
responsive to the desires of treatment planners? Keywords:
Prioritized optimization; isolated tradeoffs.
- Eva Lee, Ph.D. OR. Computational Optimization techniques applied to
brachytherapy, external beam and IMRT treatment planning. Topics related to
the application of computational optimization to treatment planning. Issues
include translating treatment planning parameters into mathematical models;
incorporation of dose-volume constraints, dose homogeneity, TCP and NTCP
criteria into treatment models; improvement of TCP via MRS-image-guided
dose escalation; computerized parallel plan generation with multiple objectives
and ranking schemes.
3:30 -- Break (15 minutes)
3:45 -- 3 talks (each 20 min + 10 min Q&A)
- Ron Rardin, Ph.D. OR. Mixed integer approaches to external
beam radiotherapy and IMRT.
A mixed integer programming approach to near-optimal
radiotherapy planning assuming a fixed set of candidate beam
angles and dose linear in beam intensity. Minimum tumor dose
is maximized subject to explicit restrictions on tumor-dose
homogeneity, and both maximum and dose-volume limits on
healthy tissue dose. Constraint-generation schemes sharpen
bounds from continuous relaxations, and column-generation is
employed to compute desirable IMRT patterns.
- Lei Xing, Ph.D. Physicist. Fast optimization techniques
and the selection of beam angles
Incorporating prior knowledge into
optimization of IMRT dose and beam configuration.
Keywords: a priori information; heuristics; selecting the best
set of beam directions.
- Julian Rosenman, M.D., Ph.D. What the oncologic community
has gained from optimization in radiotherapy, what more it
needs, and how it evaluates approaches to optimization.
Has optimization processes provided better treatment for some
patients? Has it led to inferior treatments? What level of
quality assurance is needed for optimization processes?
Are physicians prepared to be involved? What is required of
them? Can the oncologic community tell whether optimization
processes provide a benefit?
5:15 -- discussion panel (30 minutes)
Dinner (on your own)
6:30pm --
Attendees are encouraged to deposit slips with questions, opinions, suggestions
into the Suggestion Box by 9pm for
consideration in Saturday's discussions.
Feb 9: Morning Session: "Brainstorming/discussion -- The Present and the Future"
7:30am -- Continental breakfast
8:00am -- 4 panels (each 25 minutes)
Panel Discussion
Comments and discussion for each presentation session on Friday, including
suggestions slips.
Group 1: Reviewing talks 1-3
Panelists: Clifford Chao, M.D., Laura Dawson, M.D., Isaac Rosen, Ph.D., Joseph F. Pekny, Ph.D.
Model formulation I: stating constraints and objectives in treatment
planning.
Does the planning problem need reformatting?
Can better solutions be defined and recognized?
Will mathematically superior solutions improve medical practice,
policy or research?
Are there improvements needed to the solution process in addition
to the solutions themselves?
Group 2: Reviewing talks 4-6
Panelists: Mary Austin-Seymour, M.D.,
Yunping Zhu, Ph.D.,
Tim Fox, Ph.D.,
Stephen Pollock, Ph.D.
Model formulation II: Multiple objectives.
Are there multiple objectives?
Do multiple objectives apply to a single physician, or do they
appear only when polling the community?
How do we deal with
the inherently conflicting objectives in radiation
therapy?
How should planning solutions be presented if there are multiple
objectives?
How can treatment delivery methods be compared if there are
multiple objectives?
How can protocols be written to reflect multiple objectives
across the community?
Break -- 10 minutes
Group 3: Reviewing talks 7-9
Panelists: Mack Roach, M.D.,
Andrzej Niemierko, Ph.D.,
Illya VaShun Hicks, Ph.D.,
Yin Zhang, Ph.D.
What is a solution? How do we define solution accuracy? Model Formulation II: How
should volumes (and surfaces) and other clinical parameters be modeled for the purpose
of RT optimization? Can operations research methods prioritize obstacles to dose
escalation? Is ''almost sparse'' a useful concept? Do we need new computational tools
when the models become too complex?
Group 4: Reviewing talks 10-12
Panelists: Christopher M. Rose, M.D.,
Cedric X. Yu, DSc,
Shane Henderson, Ph.D.,
Ariela Sofer, Ph.D.
What tests are demanded?
Can operations research techniques provide objective methods to measure
claimed gains from new treatment modalities and resources?
Can operations research methods improve treatment or reduce costs?
How should we prepare for optimization given non linear dose effects?
Break -- 10 minutes
10:00 am -- Open floor discussion and concluding remarks
Open floor discussion led by the three co-chairs: Deasy,
Langer, and Lee (40 minutes) based on submitted audience statements
Concluding Remarks by NCI and NSF (Deye, Mahoney, Rardin)
12:00pm -- Workshop adjourns and departure for airport
|
|
