📝 SummarXiv

Abstract

Background: Current mathematical quantification methods for beam symmetry are highly sensitive to noise, especially in beam profiles with significant variation. Purpose: This study evaluates the accuracy of standard radiotherapy beam symmetry metrics and compares them to a proposed cross-correlation-based metric called the Correlative Symmetric Index (CSI), as well as the Structural Similarity Index (SSIM). We aim to demonstrate that CSI is less susceptible to noise than traditional methods. Methods: Simulated non-symmetric beam profiles with similar left and right areas were analyzed using both standard and proposed symmetry metrics. To test the robustness of each method to noise, a noisy non-symmetric beam profile was also generated. Measured beam profiles at various depths in a water tank were used to compare the performance of each symmetry metric under realistic clinical conditions. Results: In the noisy, non-symmetric case, CSI and SSIM values were 0.387 and 0.5401, respectively. Traditional metrics such as the Point Difference Quotient (PDQ) and area-based symmetry yielded values of 0.312 and 0.400. The percentage change between the non-symmetric (no noise) and non-symmetric (with noise) cases for PDQ, area-based symmetry, SSIM, and CSI were 11.1%, 5.6%, 16.25%, and 1.43%, respectively. For clinically measured symmetric profiles, all metrics produced values above 0.9. Conclusion: The Correlative Symmetric Index (CSI) demonstrates greater robustness to noise than both pointwise and area-based symmetry methods. CSI provides a reliable measure for quantifying beam symmetry, particularly under noisy conditions.