Application Of Radiomics In Predicting The Efficacy Of Immunotherapy For Solid Tumors

The first part Radiomics nomogram for predicting initial benefits from immune checkpoint inhibitor in advanced non-small-cell lung cancerObjective: To assess the value of a radiomics nomogram for predicting early treatment response prior to immune checkpoint inhibitor(ICI)for advanced non-small-cell lung cancer(NSCLC)patients.Methods: This retrospective study enrolled 64 patients with advanced NSCLC who underwent ICI monotherapy with pembrolizumab or nivolumab.All target lesions were delineated on CT imaging with and without contrast-enhanced.For each CT sequence,a total of 1967 radiomics features were extracted from the whole target lesion.Responses were evaluated by CT according to immune RECIST(i RECIST)standard and patients were divided into a responder group and a non-responder group.A radiomics signature was developed with features selected by the least absolute shrinkage and selection operator(LASSO)algorithm.Then,the predictive performance of radiomics signature was validated in an independent chemoimmunotherapy cohort of 33 patients.The clinical data included age,sex,smoking status,pleural effusion,metastasis status of pleural,brain,adrenal,bone,liver,and neck lymphonodus.Finally,a nomogram integrating the clinical factor and radiomics signature was established to predict treatment response.The area under the receiver operating characteristic curve(AUC),calibration and decision curves analysis(DCA)were used to assess the prediction ability of the nomogram.Results: Ninety-seven patients were included in the analyses,and 3 of them experienced pseudoprogression.A 5-feature radiomics signature for predicting response was developed and validated in three sets.The radiomics nomogram based on clinical factor(bone metastasis)and radiomics signature provides the best results for predicting response to ICI.The AUCs of 0.878(95% confidence interval [CI]0.812-0.944)in training set,0.756(95% CI 0.615-0.897)in testing set and 0.790(95%CI 0.660–0.921)in chemoimmunotherapy validation set.The nomogram provides more net benefit by DCA curve than clinical factor and radiomics signature alone.Conclusion:1.CT based radiomics could predict early treatment response to immune checkpoint inhibitors in patients with advanced non-small cell lung cancer.2.Among clinical factors,patients with bone metastasis receive less effective ICI therapy.3.The nomogram model based on the combination of radiomics signature and clinical factors improves the prediction ability of early treatment response of immune checkpoint inhibitors for patients with advanced non-small cell lung cancer,and this method could provide strong support for clinical medication decision.The second part Pre-treatment CT-based radiomics signatures as predictors of atypical responses to immunotherapy across tumor typesObjective: To develop and validate radiomic signatures to predict atypical responses to immune checkpoint inhibitor(ICI)in cancer patients.Methods: In this retrospective multicenter study,we collected 463 patients with mutil-types of malignancies receiving ICI.According to the criteria of immune response evaluation criteria in solid tumors(i RECIST),the atypical responses in immunotherapy were identified,including pseudoprogression(Ps P)and hyperprogressive disease(HPD).A subgroup of standard progression disease(s PD)in2018 was also involved in this study.Three prediction radiomics signatures were built to distinguish Ps P,HPD,and s PD.A total of 107 features were extracted from peritumoral and intratumoral region based on pre-treatment CT imaging separately.The least absolute shrinkage and selection operator(Lasso)algorithm was used for feature selection,multivariate logistic analysis was used to develop radiomics signatures.Clinical data included age,sex,previous treatment times,single or combined treatment,brain metastasis,bone metastasis,lung metastasis or liver metastasis.The area under the receiver operating characteristic(ROC)curve(AUC)and accuracy were applied to evaluate the performance of the radiomic signatures.Decision curve analysis(DCA)was implemented to evaluate the clinical utility.Results: The incidence rate(10.44%)of HPD is slightly higher than that(8.25%)of Ps P in the whole cohort.No significant difference was found when compared in terms of clinical characteristics of Ps P,HPD,and s PD.Radiomics signatures based on combined peritumoral and intratumoral images outperformed those from either of the two regions alone,yielding an AUC(or accuracy)of 0.834(0.827)for Ps P vs.HPD,0.923(0.868)for Ps P vs.s PD,0.959(0.894)in the training cohort.Compared with intratumoral images,features derived from peritumoral images yielded higher area under AUCs in both Ps P vs.HPD and Ps P vs.s PD.DCA showed that the radiomics signatures produced net benefit.Conclusion:1.Pretreatment CT radiomics helps to predict atypical responses in immunotherapy for different tumor types.2.The radiomics features of the peritumoral region can help to predict the occurrence of Ps P.3.The combination of intratumoral and peritumoral radiomics features better reflects the heterogeneity and microenvironment of the tumor.