Functional Imaging Technology In Early Prediction Of The Response Of Neoadjuvant Chemotherapy For Breast Cancer

Background:Neoadjuvant chemotherapy（NAC）can make inoperable locally advanced breast cancer become operable and let patients achieve long-term survival.However,the heterogeneity of breast cancer makes NAC risky,and how to early predict the response of NAC is a popular research issue.Functional imaging techniques can detect tumor neovascularization and metabolic changes that occurs earlier than morphological changes.This study explores the feasibility of functional imaging techniques such as breast-specific gamma imaging（BSGI）,diffusion-weighted magnetic resonance imaging（DWI）and dynamic contrast-enhanced magnetic resonance imaging（DCE-MRI）to early predict the response of NAC.Methods:The research was divided into 4 parts.In the first part:analyzed 80 cases of patients who completed surgery after NAC and accepted MRI and/or BSGI examinations at three time points:before NAC,before the third or fifth cycle of NAC and before surgery,focusing on tumor normal ratio（TNR,the parameter of BSGI）,apparent diffusion coefficient（ADC,the parameter of DWI）and maximum slope of increase(Slope_max,the parameter of DCE-MRI)to early predict the response of NAC.The main predictors were pathologic complete response（p CR）and Miller-Payne（MP）grading.In the second part:compared the effectiveness of BSGI and DWI in predicting the response of NAC of different molecular types of breast cancer.In the third part:downloaded the MRI images and pathological results of the QIN-BREAST data set from TCIA database,calculated the ADC of tumor area and evaluated its effectiveness in predicting p CR.In the fourth part:compared the MRI or/and BSGI from our hospital to evaluate the consistency of clinical complete response（c CR）and p CR.Result:In the first part:ΔTNR（Δrefers to the difference value）,ΔTNR%,ADC1 andΔADC were significantly correlated with p CR and MP grading after NAC respectively（p<0.05）,with area under curve（AUC）≥0.7.ADC1 had the largest AUC,which was0.810.ΔTNR,ΔTNR%,ADC1,ΔADC,andΔADC%were fitted to the MP grading by ordinal polytomous logistic regression model.The regression coefficients were tested by chi-square,and the results of p<0.05 were obtained.Constructed a Logistic regression model ofΔTNR%combined with ADC1 to predict p CR and drew ROC curve,AUC=0.865.ΔTNR,ΔTNR%,ADC1,andΔADC obtained before the third cycle of neoadjuvant chemotherapy did not show a significant difference in predicting p CR compared with the results obtained before the fifth cycle,p≥0.05.In the second part:AUC of△TNR and△TNR%predicting the response of triple-negative breast cancer NAC were greater than that of Luminal and HER2-amplified breast cancer,while AUC of ADC1 and△ADC predicting the response of HER2-amplified breast cancer better than that of Luminal and triple negative breast cancer.In the third part:the analysis of the DWI images of QIN-BREAST data set were consistent with the results of our hospital.ADC could early predict p CR,with AUC≥0.7.In the fourth part:the preoperative MRI and BSGI assessment of c CR were consistent with the postoperative pathological assessment of p CR.The sensitivity of MRI（94.12%）is higher than that of BSGI（87.50%）,and the specificity of BSGI（56.25%）is higher than that of MRI（33.33%）.Conclusion:Functional imaging technology BSGI’sΔTNR,ΔTNR%and DWI’s ADC1 andΔADC can early predict the response of breast cancer NAC.Before the third NAC cycle,ΔTNR%and ADC1 predict best;BSGI is found to have the highest accuracy in predicting triple-negative breast cancer,while DWI predicts HER2-amplified breast cancer better.The c CR assessed by MRI and BSGI has a high consistency with the p CR,suggesting that MRI and BSGI can be combined to early predict c CR and p CR.