Multidimensional Information And Radiomics Fusion Model For Predicting Radiation Pneumonia In Thoracic Esophageal Cancer

BackgroundEsophageal cancer is one of the most common malignant tumors in China,and radiotherapy(RT)plays a key role in both preoperative neoadjuvant chemoradiotherapy for operable locally advanced or radical chemoradiotherapy for inoperable locally advanced esophageal cancer.Radiation pneumonitis(RP)is one of the main toxicities of thoracic RT.Severe RP not only lacks effective treatment but also directly affects the quality of life and prognosis of patients.Therefore,achieving early assessment and prediction is critical to reduce the incidence of RP and maximize therapeutic gain.At present,the RP risk assessment of patients with esophageal cancer receiving RT is mainly measured by lung dose-volume parameter in clinical practice.However,with the progress of RT technology,the mode of RT for esophageal cancer has changed from the traditional two-dimensional to the current three-dimensional RT.This includes three dimensional conformal radiation therapy(3 DCRT)and intensity modulated radiation therapy(IMRT).The changes in RT techniques have led to changes in the dose distribution of lung tissue,so the traditional dosimetric parameters based on the two-dimensional era do not seem to be suitable for the current model.In addition,the clinical characteristics and dosimetry parameters influencing the occurrence of RP are mostly derived from the studies after RT of lung cancer.The heterogeneity,treatment differences and anatomical structure differences between lung cancer and esophageal cancer limit the transformation of factors affecting the occurrence of RP between them.Therefore,there is an urgent need for an independent reference standard that can predict the factors affecting the occurrence of RP in esophageal cancer.Computed tomography(CT)is an indispensable part of disease diagnosis,which is often used in clinical diagnosis because of its noninvasive method to acquire tissue features.The first step in the diagnosis of RP is the change of lung tissue density by CT imaging,which is manifested by the infiltration and exudation of inflammatory cells and fibrin deposition in the alveolar tissue after the lung tissue receives different degrees of radiation irradiation.Therefore,it is of great significance to explore and reveal the correlation between early changes in lung density and radiation dose in RT as well as the relationship between such correlation and the occurrence of RP for the implementation of individualized treatment and prevention of RP.CT imaging assessment of diseases mainly relies on qualitative characteristics,such as tissue density and composition,tissue distribution and anatomical relationship,etc.However,these qualitative phenotypic descriptions cannot meet the requirements of today's personalized precision medicine.In recent years,information science centered on big data has triggered a revolution in medical thinking and methods,among which radiomics has the ability to capture tumor phenotypes and tissue texture features by automatically extracting image features and converting image data into high-dimensional mining feature space.To some extent,the early diagnosis,curative effect monitoring and adverse reaction prediction of tumor have been realized.Previous studies have found the relationship between some changes in second-order or higher-order eigenvalues of imaging in chest RT and the occurrence of RP.Unfortunately,due to the limitations of detection techniques or other factors,no value-based prediction model of RP for clinical practice has been established.In this study,we extracted the imaging characteristics of lung cone-beam CT(CBCT)in different time periods from RT,and combined with the clinical characteristics and lung dose-volume parameters to establish the nomogram,so as to realize the value and significance of RP individualized prediction in RT practice of esophageal cancer.Part I:Analysis of related factors of radiation pneumonia after radiotherapy for thoracic esophageal cancerObjective:The clinical characteristics,radiotherapy techniques(RTT),lung dose-volume parameters(LDVP),RT and chemotherapy were analyzed to explore the factors affecting the incidence of RP in thoracic esophageal cancer,so as to provide reference for the formulation of RT for esophageal cancer.Methods:In this study,247 cases of esophageal cancer patients who received RT in our hospital in recent years were collected retrospectively.All patients had inoperable or unresectable thoracic esophageal cancer and received radical RT or chemoradiotherapy(CRT).Through the analysis of many clinical characteristics,three different RTT,LDVP,chemotherapy,and then to univariate and multivariate analysis of these factors,the receiver-operating characteristic curve(ROC)to verify different levels of the diagnosis efficiency of RP.Results:Among 247 patients,118 cases(47.8%)were ? grade 1 RP,54 cases(21.9%)were ? grade 2 RP,and 17 cases(6.9%)were ? grade 3 RP.Univariate analysis showed that V5-V40 and MLD in double lungs were correlated with the occurrence of grade ?1 RP(Z=-5.802?-4.306,P<0.05),grade ?2 and grade ?3 RP(F=0.057?11.616,0.087?3.392,P<0.05).Tumor target volume(GTV),planned target volume(PTV),GTV/lung volume(%),and PTV/lung volume(%)were correlated with the occurrence of grade?1(Z=-3.377?-2.041,P<0.05)and grade ?2 RP(F=3.600?9.801,P<0.05).Smoking index(SI)>400 was significantly correlated with grade ?3 RP(?2=13.295,P<0.05),chronic obstructive pulmonary disease(COPD)was significantly correlated with grade?1 RP(?2=9.146,P<0.05),while three different RTT,chemotherapy factors,RT dose,esophageal cancer stage and location were not significantly correlated with RP.The difference was not statistically significant(all P>0.05).Multivariate analysis showed that total lung relative volume?5 Gy(V5)and total lung relative volume ?40Gy(V40)were independent risk factors for grade ?1 RP,and the optimal cut-off AUC values were 55.74%and 4.13%,respectively.Mean lung dose(MLD)was an independent risk factor for grade ?2 RP,and the optimal cut-off value was 11.91 Gy.Double lung V5 was an independent risk factor for grade?3 RP,with an optimal cut-off value of 57.60%.Smoking index>400 was an independent risk factor for grade ?3 RP(Wald=5.964,P<0.05),and COPD was an independent risk factor for grade ?1 RP(Wald=6.110,P<0.05).Conclusions1.Regardless of the use of 3D-CRT,IMRT or TOMO,there is no influence on the occurrence of RP in different grades.2.Different combination methods of RT and chemotherapy,different chemotherapy regimens,and different tumor stages and locations are not factors affecting the occurrence of RP.3.Lung dose-volume parameters V5,V40 and MLD were closely related to the occurrence of RP among the corresponding grades.SI>400,COPD were an important risk factor for grade ?3 RP and ?1 RP,respectively.Part ? Correlation between lung density changes under different dose gradients and radiation pneumonitis--based on an analysis of computed tomography scans during esophageal cancer radiotherapyObjective:To assess the relationship between different doses of radiation and lung density changes and to determine the ability of this correlation to identify esophageal cancer patients who develop RP and the occurrence time of RP.Methods:A planning computed tomography(PCT)scan and a re-planning computed tomography(rPCT)scan were retrospectively collected for each of 103 thoracic segment esophageal cancer patients who underwent RT.The isodose curve(IC)was established on the planning CT with an interval of 5Gy,which was used as the standard for dividing different gradient doses.Rigid registration(RR)was conducted between PCT and rPCT,and the mean lung CT value(HU)between different doses gradients was automatically obtained by the software system.The density change value(?HU)was the difference of CT value between each dose gradient before and after treatment.The correlation between ?HU and the corresponding dose was calculated as well as the regression coefficients.Additionally the correlation between ?HU and the occurrence and time of RP(<4weeks,4?12 weeks,>12weeks)was calculated.Results:The radiation dose and ?HU was positively correlated,but the correlation coefficient and regression coefficient were lower,0.261(P<0.001)and 0.127(P<0.001),respectively.With the increase of radiation dose gradient,?HU in RP?2 group was higher than that in RP<2 group,and there was significant difference between two groups in ?HU20-25,?HU25-30,?HU30-36,?HU35-40,?HU40-45,?HU45-50(p<0.05).The occurrence time of RP was negatively correlated with the degree of?HU(P<0.05),with a high correlation coefficient(Y=Week actual value-0.521,P<0.001)(Y=Week grade value-0.381,P=0.004)and regression coefficient(Y=Week actual value-0.503,P<0.001)(Y=Week rating value-0.401,P=0.002).Conclusions:1.This study confirmed that for most dose gradient interval in RT for esophageal cancer,lung dose and corresponding lung density were positively correlated with the increase of lung irradiation dose,although the correlation was low.2.After ? 20Gy dose of lung irradiation(?Hu20-50),lung density changes significantly increased,and patients with such significant lung dose-density changes predicted an increased risk of grade ?2 RP.3.For patients with RP,the more obvious ?HU,the earlier the occurrence of RP,there was a significant negative correlation between them.Therefore,the time when patients may develop RP can be evaluated by analyzing the change range of lung density during RT,so as to provide help for early intervention of RP.Part ? A nomogram model based on cone-beam CT radiomics analysis technology for predicting radiation pneumonitis in esophageal cancer patients undergoing radiotherapyPurpose:In the first part of the study,it was confirmed that the risk prediction of RP in esophageal cancer was mainly achieved through the assessment of lung dose-volume parameters(such as V5 and MLD),but the predictive efficacy of these indicators needed to be further improved.In this part,we quantitatively analyzed the characteristics of cone-beam computed tomography(CBCT)radiomics in different periods during radiotherapy(RT),and then built a novel nomogram model integrating clinical features and dosimetric parameters for predicting RP in patients with esophageal cancer.Methods:This study included two independent study queues:primary cohort and validation cohort.Clinical characteristics and dosimetric parameters of 96 patients with esophageal cancer and CBCT images of each patient in three different periods of RT were obtained.The images were segmented using both lungs as the region of interest(ROI),and 851 image features were extracted.The least absolute shrinkage selection operator(LASSO)was applied to identify candidate radiomics features,and logistic regression analyses were applied to construct the Rad-score.The optimal period for the Rad-score,clinical features and dosimetric parameters were selected to construct the nomogram model,then the receiver operating characteristic(ROC)curve was used to evaluate the prediction capacity of the model.Calibration curves and decision curves were used to demonstrate the discriminatory and clinical benefit ratios,respectively.Results:When the three-time periods were modelled,the first period was better than the others.In the primary cohort,the area under the ROC curve(AUC)was 0.700(95%confidence interval(CI)0.568-0.832),and in the independent validation cohort,the AUC was 0.765(95%CI 0.588-0.941).V5,MLD and tumor stage were independent predictors of RP and were finally incorporated into the nomogram.In the nomogram model that integrates clinical features and dosimetric parameters,the AUC in the primary cohort was 0.836(95%CI 0.700-0.918),and the AUC in the validation cohort was 0.905(95%CI 0.799-1.000).The nomogram model exhibits excellent performance.Calibration curves indicated a favorable consistency between the nomogram prediction and the actual outcomes.The decision curve exhibited satisfactory clinical utility.Conclusion:1.The radiomics features based on lung CBCT images can effectively predict radiation pneumonia in patients with thoracic esophageal cancer,which stand for a high-risk population.2.The comprehensive nomogram combined lung CBCT Rad-score in early RT,clinical features and lung dose-volume parameters showed good predictive performance in both the primary cohort and the validation cohort.It can be used as a potential tool for individualized management of radiation pneumonia in clinical practice.