Prognosis Prediction Of Non-Small Cell Lung Cancer Patients Receiving Targeted Therapy Based On Radiomics

Lung cancer answers for the highest cancer incidence around the world,and non-small cell lung cancer(NSCLC)accounts for about 85% of lung cancer.The use of tyrosine kinase inhibitors(TKI)for targeted therapy of NSCLC patients with tumor-driven genes can rapidly improve symptoms of patients.However,patients often develop resistance to TKI therapy,thus causing very poor prognosis.In view of the above clinical problems,this paper conducts an in-depth study on the prognosis prediction model of targeted therapy for NSCLC patients based on radiomics.This article discussed the current studies on prognostic prediction of NSCLC patients receiving targeted therapy at home and abroad in detail,and analyzed the advances of radiomics in auxiliary diagnostic model construction.Radiomics methods combining artificial intelligence algorithms and medical image-based diagnosis were discussed.And the feasibility of applying radiomics to the prognosis prediction research in NSCLC patients receiving targeted therapy was well demonstrated.A machine learning-based prognosis prediction model was constructed by Cox proportional hazards regression.Firstly,the image processing algorithm was designed to extract quantitative features from computed tomography(CT)images of NSCLC patients.Then,feature correlation analysis and the least absolute shrinkage and selection operator(LASSO)algorithm were adopted for feature reduction.Finally,a prognosis prediction model was built by Cox proportional hazards regression with the optimal feature set fitted.Model evaluation by KaplanMeier survival curve analysis proved that the model could accurately predict the prognosis of NSCLC patients receiving targeted therapy.In addition,the concordance index of the model on the test set is up to 0.717,which is more accurate than the previous study.A deep convolutional neural network(DCNN)based prognostic prediction model was built by autoencoder and Deep Surv survival analysis network.The unsupervised convolutional autoencoder enhanced the model in ability of extracting deep learning features from medical image data with a small sample size.The DCNN model was then constructed by combining L1 regularization and Deep Surv network.After model evaluation,this paper first proved that DCNN models could successfully predict the prognosis of ALK-positive NSCLC patients receiving targeted therapy,and its prediction performance on small sample CT image data is similar to the machine learning model.Finally,it was concluded that by combining unsupervised feature learning,deep neural networks based on small sample sized data could assist doctors with auxiliary treatment decisions.