Bayesian Information-Theoretic Calibration of Radiotherapy Sensitivity Parameters for Informing Effective Scanning...

Area of Science:

• Computational Biology and Bioinformatics
• Mathematical Oncology
• Medical Imaging and Data Analysis

Background:

• Technological advancements enable comprehensive tumor growth data collection (volume, composition, vascularity).
• Significant patient variability in treatment response necessitates accurate, early inference of patient-specific parameters.
• Current sparse data collection schedules due to cost and invasiveness limit timely treatment adjustments.

Purpose of the Study:

• To determine optimal data collection times and metrics for informing tumor growth and treatment models.
• To maximize information gain about patient-specific parameters within a limited measurement budget.
• To enable early and accurate adjustment of cancer treatment protocols for improved efficacy.

Main Methods:

• Employed a Bayesian information-theoretic calibration protocol for experimental design.
• Utilized sequential data collection to maximize reduction in parameter uncertainty with each measurement.
• Developed a framework for selecting optimal metrics at each data collection point.

Main Results:

• Identified sequential data collection strategies that maximize information gain for model parameters.
• Demonstrated a framework for selecting the most informative metrics for calibration.
• Analyzed the impact of measurement budget on model predictive power using radiotherapy examples.

Conclusions:

• Bayesian information-theoretic approaches can optimize experimental design for calibrating tumor growth models.
• Strategic selection of data collection times and metrics enhances patient-specific parameter inference.
• This framework supports adaptive treatment strategies by maximizing insights from limited, costly measurements.