Predictive Value Of Pulmonary CT Texture Features For Radiation Pneumonitis

Objective: To explore the predictive value of pre-radiotherapy pulmonary CT texture features for radiation pneumonitis.Methods: The patients with lung cancer who received chest radiotherapy in the First Affiliated Hospital of China Medical University from January 2017 to March 2020 were included,and the clinical characteristics,dosimetry parameters and pre-radiotherapy pulmonary CT images of the patients were collected.We set LUNG-GTV,V5 lung tissue and V20 lung tissue of the patient’s affected side as the region of interest(ROI)in the pre-radiotherapy pulmonary CT,and use the pyradiomics package in Python to extract the original texture features.The independent influencing factors of radiation pneumonia were analyzed by univariate analysis and Logistic regression model.Results: A total of 379 patients were included in the study.According to the follow-up results,311 patients developed ≥grade 2 radiation pneumonitis(RP),and 68 patients did not develop radiation pneumonia.Univariate analysis showed age,T stage,4 texture features(glcm＿Cluster Prominence,glcm＿Cluster Shade,glcm＿Idmn,glcm＿Idn)of the ipsilateral LUNG-GTV,2 texture features(firstorder＿Energy,firstorder＿Total Energy)of the ipsilateral V5 lung tissue,and 3 texture features(shape＿Maximum2DDiameter Row,shape＿Maximum3DDiameter,gldm＿Dependence Non Uniformity)of the ipsilateral V20 lung tissue were correlated with the incidence of ≥grade 2 RP(P < 0.05).Logistic regression results showed T stage,N stage and glcm＿Imc2 of the ipsilateral LUNG-GTV were significantly related to the occurrence of ≥grade 2 RP(P<0.05).Conclusion: The T stage,N stage and glcm＿Imc2 of the ipsilateral LUNG-GTV are the independent influencing factors of ≥grade 2 RP.