Bayesian calibration, validation, and uncertainty quantification of diffuse interface models of tumor growth

Area of Science:

• Computational biology
• Mathematical oncology
• Bayesian inference

Background:

• Predicting tumor emergence, growth, and decline computationally could revolutionize cancer treatment.
• Challenges include model selection, implementation, data acquisition, and quantifying uncertainty in complex biological systems.
• Accurate tumor growth prediction is crucial for developing new therapeutic and preventative strategies against deadly cancers.

Purpose of the Study:

• To establish a systematic framework for validating computational tumor growth models.
• To develop methods for quantifying uncertainty in model predictions of tumor behavior.
• To address the inherent difficulties in predicting complex biological events like tumor growth.

Main Methods:

• Utilized a framework based on Bayesian probability.
• Focused on systematically addressing model Validation and Uncertainty Quantification.
• Employed virtual data for illustrative purposes, applying it to diffuse-interface models of tumor growth.

Main Results:

• Demonstrated a structured approach to assess the accuracy of computational models in reproducing tumor growth events.
• Provided measures for the confidence level in computer model predictions concerning tumor growth.
• Successfully applied the framework using virtual data for diffuse-interface tumor growth models.

Conclusions:

• The proposed Bayesian framework offers a robust method for validating tumor growth models and quantifying prediction uncertainty.
• This approach is essential for advancing the reliability of computational predictions in oncology.
• Improved model validation and uncertainty quantification can lead to more effective cancer treatment and prevention paradigms.