| Part 1:Predicting the efficacy of radical concurrent chemoradiotherapy for cervical cancer based on clinical dataObjective:To use statistical methods to analyze the clinical data characteristics of cervical cancer patients with different curative effect groups of radical concurrent chemo-radiotherapy,to screen out the influencing factors that can predict the remission ability after radiotherapy and chemotherapy,and to establish a prediction model based on clinical data,which can be used for clinical treatment.Provide scientific evidence.Methods:A retrospective analysis of 122 cervical cancer patients admitted to the Radiotherapy Department of the Affiliated Hospital of North Sichuan Medical College from 2017-06-01 to 2020-10-01 was performed according to the RECIST criteria.Partial remission group(PR),stable disease group SD.According to clinical data such as pathological type,stage,tumor mass diameter,parametric infiltration,uterine location,inguinal lymph node metastasis,paranoiac lymphatic metastasis and other clinical data,the effective factors affecting the remission ability after radiotherapy and chemotherapy were screened out.Results:(1)Statistical analysis showed that tumor mass diameter(F=377.70,P<0.05),parametric infiltration(F=377.70,P<0.05)had statistical significance.Age,tumor stage,pathological type,uterine location,paranoiac lymph node metastasis,and inguinal lymph node metastasis are all significant factors for the efficacy of radical concurrent chemo-radiotherapy in patients with cervical cancer.The curative effect was not statistically significant(P>0.05).(2)To establish a deep learning model based on the tumor mass diameter and parametric infiltration and the curative effect of concurrent radiotherapy and chemotherapy for cervical cancer,and use the ROC curve to test the predictive performance of the model.The results showed that the area under the ROC curve was 0.655 in macro(macro)AUC value,0.657 in micro(micro)AUC value,0.774 in sensitivity,0.361 in specificity,0.522 in precision,and 0.615 in F1 value.Conclusions:(1)Tumor mass diameter and parametric infiltration are important predictors of the efficacy of concurrent chemo-radiotherapy for cervical cancer.Follow-up treatment can be added to standard concurrent chemo-radiotherapy for these patients;Age,tumor stage,pathological type,uterine location,parailiac lymph node metastasis,and inguinal lymph node metastasis were not important predictors of the efficacy of concurrent chemoradiotherapy for cervical cancer.(2)The prediction model based on clinical data established by tumor mass diameter,parametric infiltration and curative effect of curative concurrent chemo-radiotherapy for cervical cancer has high predictive value for curative effect,and can be applied to the clinic later to provide a scientific basis for clinical treatment.Part Ⅱ:Radio-mics study based on dynamic contrast-enhanced MRI combined with DWI and T2WI model in predicting the efficacy of radical concurrent chemo-radiotherapy for cervical cancerObjective:This study aims to explore the application value of Radio-mics based on dynamic contrast-enhanced MRI combined with DWI and T2WI models in predicting the efficacy of radical concurrent chemo-radiotherapy in patients with cervical cancer,and add clinical data to construct a combined clinical+ multimedia magnetic resonance model.To study its application value in predicting the efficacy of radical concurrent chemo-radiotherapy in patients with cervical cancer,and to provide a scientific basis for clinically formulating patients’ individualized treatment plans and subsequent follow-up plans in a timely manner.Methods:A retrospective analysis of 122 patients with cervical cancer who underwent radiotherapy in the Radiotherapy Department of the Affiliated Hospital of North Sichuan Medical College from 2017-06-01 to 2020-10-01,all patients underwent conventional plain scan+enhanced magnetic resonance examination before and after treatment,referring to the end of the radiotherapy cycle After the efficacy evaluation,the groups were divided into two categories(CR group and non-CR group,PR and non-PR group,SD and non-SD group)and three categories(CR group,PR group and SD group).The magnetic resonance T2WI digitalis,DWIH,and DWIL images of all patients before treatment were exported from the PACS system in DICOM format,and the images were imported using the target volume delineation software Monaco,and the tumor area was delineated layer by layer to obtain the three-dimensional image information of the tumor.The images were imported into the feature analysis software 3Dslicer to extract feature values,and 1086 sets of image feature information were obtained.The high-level computer language Python is used to import data,standardize and randomize the data,and perform feature optimization to establish a predictive model.Receiver operating characteristic curve(ROC)is used to evaluate the predictive performance of the model.The clinical model,the multimedia magnetic resonance model and the clinical+ imaging model were compared to predict the curative effect,and the optimal model for curative effect prediction was obtained.Results:(1)In the DCE-MRI sequence,sag1 is the enhanced arterial phase,and sag2 and sag3 are the enhanced venous phase.The accurate value of sag2 in the dichotomous CR group and SD group is higher than that of sag3,and the accurate value of DWIL in the CR group is higher than that of DWIH.The accurate value of DWIH in SD group was higher than that of DWIL.The AUC of sag2 and sag3 sequences in the binary CR group was equal to 0.727,and the AUC of DWIH sequence of 0.753 was greater than that of DWIL sequence of 0.734.In the two-category SD group,the AUC of sag2 sequence of 0.908 was greater than that of sag3 sequence of 0.800,and the AUC of DWIH sequence of 0.850 was greater than that of DWIL sequence of AUC 0.742.(2)The accuracy of combination 1(T2WI+DCE-MRI)and combination 4(T2WI+DCE-MRI+DWI)in the two-category CR group and SD group were higher than those of combination 2(T2WI+DWI)and combination 3(DCE-MRI)+DWI).In the CR group,the AUC of combination 1 was 0.848,the AUC of combination 2 was 0.809,the AUC of combination 3 was 0.833,and the AUC of combination 4 was 0.873.The AUC of combination 1 and combination 4 was higher than that of combination 2 and combination 3,and combination 4(T2WI+DCE-MRI+DWI)was the highest.In SD group,the AUC of combination 1 was 0.933,the AUC of combination 2 was 0.928,the AUC of combination 3 was 0.939,and the AUC of combination 4 was 0.956.The AUC of combination 1 and combination 3 was higher than that of combination 2 and combination 4,and combination 4(T2WI+DCE-MRI+DWI)was the highest.(3)The accurate value of sag2 is higher than that of sag3 for three categories,and the accurate value of DWIH is higher than that of DWIL.sag2 sequence macro(macro)AUC 0.814,micro(micro)AUC 0.817,sag3 sequence macro(macro)AUC 0.670,micro(micro)AUC 0.660,sag2 sequence macro(macro)AUC and micro(micro)AUC are higher than sag3 sequence.DWIL sequence macro(macro)AUC 0.728,micro(micro)AUC 0.731,DWIH sequence macro(macro)AUC 0.757,micro(micro)AUC 0.763,DWIH sequence macro(macro)AUC and micro(micro)AUC are higher than DWIL sequence.(4)The accuracy of the three-category combination 1 and combination 4 is higher than that of combination 2 and combination 3.Combination one macro(macro)AUC 0.886,micro(micro)AUC 0.877,combination two macro(macro)AUC 0.795,micro(micro)AUC 0.790,combination three macro(macro)AUC 0.844,micro(micro)AUC 0.842,combination four Macro(macro)AUC 0.894,micro(micro)AUC 0.898.The macro AUC and micro AUC of combination 1 and combination 4 were higher than those of combination 2 and combination 3,and combination 4(T2WI+DCE-MRI+DWI)was the highest.(5)In the two-category CR group,the AUC of clinical data+combined one model was 0.867,and the AUC of clinical data+combined four model was 0.885.In the SD group,the AUC of clinical data+combined one model was 0.951,and the clinical data+combined four model AUC was 0.962.Therefore,the AUC value of clinical data+combination one model in CR group was higher than that of clinical data+combination four model,while the AUC value of clinical data+combination four model in SD group was higher than that of clinical data+combination one model.(6)Three-category clinical data+combined one model macro(macro)AUC 0.891,micro(micro)AUC 0.882,clinical data+combined four clinical data model macro(macro)AUC 0.926,micro(micro)AUC 0.931.Therefore,the macro(macro)AUC and micro(micro)AUC values of clinical data+combination four(T2WI+DCE-MRI+DWI)were higher than those of clinical data+combination one(T2WI+DCE-MRI).(7)The accuracy and AUC value of the clinical data+multimodal magnetic resonance imaging omics model were higher than those based on dynamic contrast-enhanced MRI combined with DWI and T2 omics models were higher than the clinical data model.The research carried out the Delong test on the ROC curve,P<0.05,and the AUC differences were all significantly different.Therefore,the clinical data+multimodal magnetic resonance imaging omics model is better than the clinical data model because of the combination of dynamic enhanced MRI combined with DWI and T2 omics model for the prediction of efficacy of radical concurrent chemoradiotherapy for cervical cancer.Conclusion:Clinical data(tumor mass diameter(>40mm)+parametrial infiltration)combined with multimodal magnetic resonance imaging combination model(T2WI+DCE-MRI+DWI)compared with simple clinical data model and multimodality The MRI omics model has more advantages in predicting the efficacy,and the efficacy prediction efficiency is:the clinical data+multimodal MRI omics model is better than the omics model based on dynamic contrast-enhanced MRI combined with DWI and T2WI is better than the clinical data model. |
PDF Full Text Request