SU-E-T-221: Evaluation of Technology Using Probabilistic Decision Models.

Area of Science:

• Medical Physics
• Decision Analysis
• Radiotherapy Technology Evaluation

Background:

• Medical physicists face complex decisions when selecting clinical technology, often involving uncertain variables.
• Probabilistic decision models offer a robust framework for managing these uncertainties.

Purpose of the Study:

• To present the principles of constructing probabilistic decision models.
• To provide practical examples of these models in image-guided radiation therapy (IGRT), intensity-modulated radiation therapy (IMRT), and proton therapy.

Main Methods:

• Influence diagrams, a type of directed acyclic graph, were used to model variables, uncertainties, and decision outcomes.
• Bayesian probability calculus was employed to propagate probabilities and update prior beliefs with evidence.
• A specific influence diagram evaluated brain tumor treatment comparing x-ray IMRT and proton therapy, incorporating data on tumor control probability (TCP), normal tissue complication probability (NTCP), and patient motion.

Main Results:

• Critical variables impacting treatment decisions were identified through sensitivity analyses.
• The significance of imaging, irrespective of radiation modality, was highlighted.
• The choice of conditional probability parameters influenced the ranking of treatment alternatives.

Conclusions:

• Rigorous probabilistic frameworks are essential for sound decision-making in radiotherapy technology selection.
• Influence diagrams provide a powerful tool for comparing irradiation modalities, as demonstrated in the brain tumor treatment example.
• Decisions based on unstated assumptions or incorrect inference can be mitigated by using structured probabilistic models.