Purpose: The aim of the work was to create a predictive model of the intrafraction motion variability in the superior/inferior (SI) direction applicable in margin delineation of internal target volume.
Methods: Log file analysis from online respiratory tumor tracking was performed in 145 patients. We used univariate analysis to assess the suitability of the variables recorded during data collection. Data from 100 patients were selected for model creation, and the remaining data were used for model cross validation. Categorical variables included sex (male / female), origin of lesion (metastasis / primary), lungs (left / right), combination of sex and origin. Continuous variables included mean motion along the SI axis (mm), GTV size (mm3), and localization of the target in the lung tissue: individual partial distances (mm) and relative target location (0–1) in the SI, AP, and LL directions. Backward elimination was used in the creation of the model (using p <0.05 as the criterion of statistical significance).
Results: A model was created in the following form: Intrafraction variability SI (mm)= -0.82+0.003∙Ant+0.38∙TPSI+0.006∙Mdwhere Ant is the perpendicular distance from the center of the lesion to the outermost edge of the lung in the Anterior direction (mm), TPSI is the mean Target Motion in the SI direction (mm), and Md is the length of the perpendicular line from the center of the target to the vertebra.The mean prediction error was 0.55 ± 0.48 mm. A prediction error exceeding 1 mm was recorded for five targets (11%). Predicted values highly correlates with real observed values (R=0.89, p<0.001).
Conclusion: The developed model exhibits a good ability to estimate intrafraction variability. We propose patient specific margin expansion in the SI direction based on the model output after the impact of the ITV margin expansion on dose distribution will be further tested.