A Individualized Research To Adjust Radiotherapy Dose Of Lung Cancer Patients Based On Radiomics Model

BACKGROUND: Radiomics can be used for the diagnosis and treatment of diseases fast and non-invasively.Radiotherapy is one of the main methods to treat lung cancer.Its therapeutic effect depends on the radiosensitivity of tumor cells,and the lung cancer histological type is related to the radiosensitivity.Therefore,it is very important to diagnose the histological type of lung cancer and to evaluate the radiosensitivity of lung cancer patients with Radiomics.PURPOSE: Development a radiomics prediction model for diagnosing lung adenocarcinoma,development a radiomics model for tumor volume rate prediction for lung cancer patients after radiotherapy,and development a radiomics model for individualize radiotherapy dose for lung cancer patients.METHODS: 1.Diagnose lung adenocarcinoma.This was a historical cohort study,three independent lung cancer cohorts included.One cohort was used to evaluate the stability of radiomics features,one cohort was used to feature selection,and the last was used to construct and evaluate classification models.The research is divided into four steps: region of interest segmentation,feature extraction,feature selection,and model building and validation.The feature selection methods included the intraclass correlation coefficient(ICC),ReliefF coefficient,and Partition-Membership filter.The performance metrics of the classification model included accuracy(Acc),precision(Pre),area under curve(AUC)and kappa statistics.2.Two independent lung cancer patient cohorts were analyzed,ICC,ReliefF,and Pearson’s coefficients were used for feature selection.The prediction features includes clinical features and radiomics features.Logistic regression used to build a prediction model.The performance metrics includes Acc,Pre,Recall and AUC.3.Individualize radiotherapy dose.Two independent lung cancer patient cohorts were analyzed retrospectively for feature selection and model construction.The performance indexes of feature scoring standard and classification model are the same as those in the first work.RESULTS: 1.In the first study,The 10 features(First order shape features:Sphericity and Compacity,Grey-Level Run Length Matrix: Short-Run Emphasis,Low Gray-level Run Emphasis,and High Gray-level Run Emphasis,Grey LevelCo-occurrence Matrix: Homogeneity,Energy,Contrast,Correlation,and Dissimilarity)showed the most stable and classification capability.The 6 classifiers,Logistic regression classifier(LR),Sequence Minimum Optimization algorithm,Random Forest,KStar,Naive Bayes and Random Committee,have great performance both on the train and the test sets,and especially LR has the best performance on the test set(Acc=98.72,Pre=0.988,AUC=1,and kappa=0.974).2.In the second study,3 clinical features and 3 radiomics features are selected.LR model has a good performance on test set(Acc,Pre,and AUC>85%).3.In the third study,10 radiomics features are selected,including maxValue,Skewness,Compacity,Sphericity,Energy_Grey Level Co-occurrence Matrix,Contrast_Grey Level Co-occurrence Matrix,Dissimilarity_Grey Level Co-occurrence Matrix,Low Gray-level Run Emphasis_Grey-Level Run Length Matrix,Short-Zone Emphasis_Grey-Level Zone Length Matrix,and Zone Length Non-Uniformity_Grey-Level Zone Length Matrix.Six machine learning classifiers have good classification ability on the test set(AUC > 0.90),among which the minimum sequence optimization algorithm has the best performance(ACC = 96.9%,pre = 1.00,AUC = 0.97,kappa = 0.97)CONCLUSIONS: CT based radiomics features can identify lung adenocarcinoma and optimize the radiotherapy dose of lung cancer patients.