An organ deformation model using Bayesian inference to combine population and patient-specific data

Area of Science:

• Medical Physics
• Radiotherapy
• Computational Biology

Background:

• Data-driven organ deformation models are crucial for radiotherapy but rely on either patient-specific or population data.
• Existing models face limitations in accuracy and data requirements for predicting individual motion patterns.

Purpose of the Study:

• To develop and evaluate Bayesian deformation models that combine population and patient-specific data.
• To achieve accurate prediction of individual organ motion using fewer patient scans.

Main Methods:

• Two Bayesian deformation models were derived and applied retrospectively to rectal wall motion in prostate cancer patients.
• Coverage probability matrices (CPMs) were generated and their spatial correlations with ground truth were calculated.

Main Results:

• Bayesian models demonstrated significantly higher spatial correlation with ground truth compared to patient-specific and population-derived models.
• The models showed superior performance even with limited patient-specific scans (1-3).

Conclusions:

• The proposed Bayesian framework enables accurate individual motion prediction with reduced data requirements.
• These models offer potential for robust radiotherapy planning and evaluation, improving treatment delivery and reducing toxicity.