| خلاصه مقاله | Introduction: It is believed that medical images contain pathophysiological information that is hidden from the naked eye. Radiomics is an emerging field that converts image data into a high meaningful features. After irradiation, the tissue morphology and pathology undergo changes. In this study, we aimed to determine if MVCT-based radiomics features are sensitive to radiation dose delivery during tomotherapy procedure.
Methods: In this retrospective study, the correlation between radiomics features and radiation dose delivery during early treatment sessions was investigated. Tomotherapy MVCT images from 10 brain cancer patients, whose treatment had been completed, were selected for radiomics evaluation. During radiomics analysis a circular 2D-ROI with a constant diameter was delineated inside brain tissue within 90% isodise. The second-order radiomics features from the MVCT images during the first, fifth and tenth treatment fractions were extracted by Pyradiomics toolkits. Finally, the correlation between radiomics features value and fractions number was assessed using the Spearman correlation coefficient.
Results: A total of 22 second-order texture features were extracted. These features were divided into 7 Gy-level co-occurrence matrix (GLCM), 1 Gy-level run length matrix (GLRLM), 3 neighboring Gy-level dependence matrix (NGLDM), and 11 Gy-level zone length matrix (GLZLM). Eight radiomics features between treatment fractions showed a very high spearmen correlation coefficient (>0.75). GLCM_homogeneity (0.86), GLCM_energy (0.76), GLCM_contrast (0.79), GLCM_dissimilarity (0.79), GLRLM_RP (0.85), NGLDM_coarseness (0.79), NGLDM_contrast (0.79), and NGLDM_busyness (0.75).
Conclusion: In this study, several GLCM radiomics features showed strong correlations during early treatment fractions. Previous studies have utilized high-quality MRI or kVCT for similar analyses, however, MVCT in tomotherapy is a standard routine procedure during radiotherapy. During early treatment fractions Radiomics can evaluate treatment response and personalize treatment schedules. |