Predicting Radiotherapy Patient Outcomes with Real-Time Clinical Data Using Mathematical Modelling

Area of Science:

• Oncology
• Mathematical Biology
• Radiotherapy

Background:

• Tumor response to radiotherapy varies significantly even for similar pre-treatment conditions.
• Mathematical models can predict treatment outcomes and personalize radiotherapy fractionation.
• Model complexity and sparse clinical data limit the effective use of predictive models.

Purpose of the Study:

• To develop a compartment model for tumor volume and composition.
• To create a predictive model for tumor volume progression and associated uncertainty.
• To capture inter-patient variability using patient-specific model parameters.

Main Methods:

• A compartment model for tumor volume and composition was developed.
• Novel statistical methodology and existing clinical data were used.
• A bootstrap particle filter-like Bayesian approach was employed for training.

Main Results:

• The model, despite its simplicity, can generate diverse patient responses.
• A predictive model for tumor volume progression and uncertainty was established.
• The approach was validated against unseen data, showing predictive ability.

Conclusions:

• The developed model offers a promising tool for predicting radiotherapy response in head and neck cancers.
• This approach can potentially guide personalized treatment strategies.
• The model's predictive capabilities and limitations were demonstrated.