Inter-patient image registration algorithms to disentangle regional dose bioeffects

Area of Science:

• Medical Physics
• Radiation Oncology
• Radiotherapy

Background:

• Technological advancements in radiation therapy necessitate new methods for evaluating dose-response relationships.
• Traditional dose-volume histograms may not adequately identify all dose-sensitive normal tissue regions.
• Radiation-induced morbidity (RIM) analysis requires refined dosimetric evaluation techniques.

Purpose of the Study:

• To assess the impact of elastic image registration (EIR) algorithms on voxel-based approach (VBA) results for radiation therapy.
• To evaluate the effectiveness of VBA in identifying dose-sensitive regions associated with RIM in thoracic patients.
• To explore the integration of covariates into dosimetric statistical models for VBA.

Main Methods:

• Voxel-based approach (VBA) for local dose distribution analysis.
• Inter-patient elastic image registrations (EIRs) using Demons and B-Spline algorithms.
• Statistical modeling incorporating dosimetric and non-dosimetric variables to address the multiple comparison problem.

Main Results:

• Both Demons and B-Spline EIR algorithms successfully identified overlapping subregions with dose differences linked to RIM.
• The VBA, enhanced with EIR, proved effective in highlighting critical areas for RIM analysis.
• The inclusion of covariates expanded the potential of VBA for analyzing RIM in complex datasets.

Conclusions:

• Voxel-based approach (VBA) combined with elastic image registration (EIR) offers a robust method for analyzing radiation therapy dose distributions.
• This approach enhances the identification of regions associated with radiation-induced morbidity (RIM).
• The integration of covariates further refines VBA for comprehensive RIM analysis in radiation oncology.