Prediction Of Treatment Outcome Using Multi-parametric MRI In Locally Advanced Cervical Cancer Patients Underwent CCRT

Purpose1.To testify the possibility to predict treatment response of cervical cancer patients underwent CCRT（concurrent chemoradiotherapy）by measurement of SI（signal intensity）change on T2WI（T2-weighted imaging）.2.To monitor changes of multi-b DWI（diffusion weighted-imaging）and DCE-MRI（dynamic contrast-enhanced magnetic resonance imaging）derived parameters during CCRT,and to search for possible imaging bio markers to predict treatment outcome at ultra-early stage during treatment in cervical cancer patients.3.To investigate correlations of parameters between DWI and DCE-MRI,and to weight the application value of multi-parametric MRI in clinical practice.Methods1.We retrospectively analyzed 147 cervical cancer patients underwent CCRT who conducted conventional MRI（magnetic resonance imaging）at e0（before treatment）and e25(after the completion of 25^th radiotherapy).Correlation between changes of SI and treatment outcome were studied.We prospectively recruited 48 cervical cancer patients underwent CCRT.All patients conducted conventional MRI examinations at e0,e3(after the completion of 3^rd radiotherapy),e10(after the completion of 10^th radiotherapy)and e25.Dynamic changes of tumor size and SI at four time points were monitored.Parameters were compared between CR（complete response）and PR（partial response）group patients,in order to search for imaging biomarkers to predict treatment outcome.2.We prospectively recruited 48 cervical cancer patients underwent CCRT.All patients conducted multi-b DWI with 11 b-values（0-1200 s/mm²）at e0,e3,e10 and e25.Mono-,bi-and stretched exponential models were adopted to generate diffusion and perfusion parameters.Dynamic changes of parameters at four time points were monitored.Parameters were compared between CR and PR group patients,in order to search for imaging biomarkers to predict treatment outcome.3.We prospectively recruited 48 cervical cancer patients underwent CCRT.All patients conducted DCE-MRI examinations at e0,e3,e10 and e25.Extended Tofts linear model was adopted to generate quantitative parameters.Dynamic changes of quantitative parameters at four time points were monitored.Quantitative parameters were compared between CR and PR group patients,in order to search for imaging biomarkers to predict treatment outcome.4.Correlations between DWI and DCE-MRI derived parameters were analyzed.ROC（receiver operating characteristic curve）of different diagnostic models were compared respectively.Results1.Compared with e0,tumor size,SI and TDR（tumor to disc ratio）decreased at e25（P<0.05）.FIGO stage,lymph node metastasis andΔTDR-e25 were lower in CR group patients（P<0.05）.FIGO stage,lymph node metastasis andΔTDR-e25 were independent prognostic factors for residual tumor occurrence.2.ADC（apparent diffusion coefficient）,D（pure diffusion coefficient）,f（perfusion fraction）,and DDC（distribution diffusion coefficient）increased,whileα（intravoxel water diffusion heterogeneity）decreased during CCRT（P<0.05）.ADC showed strong correlation with D and relative strong correlation with DDC（r=0.680,0.595）.Weak correlations were observed between D^*and f,D and DDC（r=0.323,r=0.302）.ADC was higher in CR group at e0（P<0.05）.ADC and f were higher in CR group at e3（P<0.05）.ADC,D and f were higher in CR group at e10（P<0.05）,whileαwas lower in CR group compared with PR group patients（P<0.05）.ADC-e0,ADC-e3 and f-e0 could act as imaging bio markers to assess and predict treatment response at ultra-early stage during treatment for cervical cancer patients.3.DCE-MRI derived quantitative parameter K^trans（volume transfer constant）increased between e0,e3 and e10,while decreased at e25（P<0.05）.K_ep（efflux rate constant）decreased during treatment（P<0.05）.V_e（extracellular extravascular vlume fraction）rose at e25（P<0.05）.K^trans was higher in CR group at e0（P<0.05）.K^trans and V_e were higher in CR group at e10（P<0.05）.At e25,K^trans and K_ep were lower in CR group,while V_e was higher in CR group（P<0.05）.K^trans-e0、K^trans-e3 and V_e-e3 could act as imaging bio markers to assess and predict treatment response at ultra-early stage during treatment for cervical cancer patients.4.f from multi-b DWI,K^trans and K_ep from DCE-MRI were statistically correlated（r=0.681,r=0.532）.Inter-observer consistency was good among multi-parametric MRI models.DCE-MRI based model 3 showed high predictive potency in single models,with AUC（area under the curve）of 0.809.Combined model1+2+3 showed highest predictive potency in all models,with AUC of 0.903.Conclusion1.ΔTDR-e25 was an independent prognostic factor for predicting residual tumor occurrence in cervical cancer patients after CCRT treatment.The combination ofΔTDR-e25 and clinical pathological features can serve as a valuable bio marker to distinguish patients with higher possibility of residual tumor occurrence.2.Functional MRI could be applied to assess and predict treatment response for cervical cancer patients at ultra-early stage during CCRT.At e3,multi-b DWI derived parameters ADC-e0,ADC-e3 and f-e3,and DCE-MRI derived parameters K^trans-e0,K^trans-e3 and V_e-e3could assess and predict treatment response with high diagnostic accuracy.3.DCE-MRI based model3 ranked first among single models in diagnostic accuracy.Multi-parametric MRI showed highest diagnostic potency among models in treatment assessment and prediction for cervical cancer patients underwent CCRT.