Exploring the impact of dosiomics features on DQA results in breast radiotherapy using TomoDirect

Abstract

This study investigates the correlation between dosiomics features and Delivery Quality Assurance (DQA) results in TomoDirect radiotherapy treatments, aiming to enhance DQA accuracy and efficiency. Dosiomics features, such as shape characteristics, statistical properties, and texture metrics (e.g., GLCM, GLSZM), were extracted from RT Dose DICOM files of patients treated with TomoDirect on the Radixact X9 system. Regions of interest (ROI), such as the planning target volume (PTV), were used to isolate dose values for feature extraction. The DQA results were classified using gamma analysis (3%/3mm criteria), and statistical methods like correlation, group comparison, and regression analysis were applied to assess the relationship between these features and DQA outcomes. The analysis identified several key predictors of DQA success. Shape features, including surface area and object size, along with texture features like GLCM autocorrelation and GLDM high gray-level emphasis, showed significant correlations with DQA pass rates. The multivariate regression model explained 79.7% of the variance in DQA outcomes, emphasizing the potential of dosiomics features to predict DQA results. In addition, features related to dose uniformity and complexity, such as firstorder_10th Percentile and GLCM contrast, significantly impacted gamma pass rates. This study demonstrates that dosiomics can enhance the predictability of DQA outcomes in TomoDirect treatments. The identified features can support the development of predictive models to streamline DQA processes, improve treatment accuracy, and reduce manual verification efforts. Future research should explore integrating additional parameters and expanding these methods to other radiotherapy techniques and machines for broader applicability.