Dose coverage calculation using a statistical shape model-applied to cervical cancer radiotherapy.

Area of Science:

• Medical Physics
• Radiotherapy
• Computational Anatomy

Background:

• Accurate radiation therapy requires accounting for patient anatomical variations during treatment.
• Existing methods often rely on single patient datasets, limiting the evaluation of dose coverage probabilities.
• Population-based statistical shape models (SSMs) offer a promising approach to capture inter-patient variability.

Purpose of the Study:

• To develop and evaluate a comprehensive methodology for treatment simulation using a population-based SSM.
• To assess the impact of patient-specific geometric deformations on dose coverage probabilities in radiotherapy.
• To compare the probabilistic evaluation from the SSM with conventional single-dataset treatment planning.

Main Methods:

• Creation of an SSM by principal component analysis of intra-patient deformations from repeated imaging.
• Mapping patient-specific deformations to a common reference frame.
• Sampling patient deformations using weights of dominant eigenmodes.
• Testing the normality assumption of eigenmode weights.
• Individualizing the SSM with new patient data.

Main Results:

• Seven eigenmodes captured 90% of deformation variance in cervical cancer patients.
• Three eigenmodes ensured stability in simulated dose coverage probabilities.
• Normality assumption for eigenmode weights was relevant for the 20 most dominant modes (excluding the first).
• SSM individualization improved with two deformation samples from a new patient.
• Probabilistic evaluation provided insights into trade-offs compared to conventional planning.

Conclusions:

• Population-based SSMs can effectively model patient-specific geometric deformations for radiotherapy simulation.
• The proposed methodology enables a more robust evaluation of dose coverage probabilities.
• Probabilistic assessment using SSMs enhances understanding of treatment uncertainties and trade-offs.