Objective To construct and validate a breast cancer prognosis prediction model based on multimodal ima-ging and clinical feature fusion.Methods 156 breast cancer patients admitted to the hospital from January 2020 to March 2023 were prospectively selected and randomly divided into a training set(125 cases)and a validation set(31 cases)ac-cording to 8:2.Patients underwent surgical resection after neoadjuvant chemotherapy and pathological complete response(pCR)was analyzed.Before and after neoadjuvant chemotherapy,high-resolution MRI,breast ultrasound and mammogra-phy were performed.The factors affecting the prognosis of breast cancer patients(pCR after neoadjuvant chemotherapy)were analyzed,and the prognosis prediction model of breast cancer based on the fusion of multimodal images and clinical features was constructed,and the model was validated and evaluated.Results In the training set,31 cases achieved pCR,and in the verification set,7 cases achieved pCR.Logistic regression analysis showed tumor stage(OR=5.254,95％CI:2.161-12.769),Doppler echo(OR=4.909,95％CI:2.020-11.930),△ apparent diffusion coefficient(ADC)(OR=4.419,95％CI:1.818～10.741),molybdenum calcification status(OR=4.358,95％CI:1.793-10.591)were the prog-nostic factors of breast cancer patients(P＜0.05).Taking the above influencing factors as predictive variables,a diagnos-tic model was established with a nomogram.The total score of each factor ranged from 89 to 374,and the corresponding risk rate ranged from 0.07 to 0.89.The verification results of the nomogram model showed that the C-index was 0.804(95％CI:0.768-0.841),and the correction curve for predicting breast cancer prognosis was close to the ideal curve(P＞0.05).ROC curve results of the training set showed that the sensitivity of the nomogram model to predict the prognosis of breast cancer patients was 77.42％(95％CI:58.46％-89.72％),the specificity was 86.17％(95％CI:77.153％-92.14％),and the AUC was 0.856(95％CI:0.778-0.939).The ROC curve results of the validation set showed that the sensitivity of the nomogram model to predict the prognosis of breast cancer patients was 71.43％(95％CI:30.26％-94.89％),the specificity was 91.67％(95％CI:71.53％-98.54％),and the AUC was 0.872(95％CI:0.7955-0.949).Conclusion The prognosis prediction model of breast cancer based on the fusion of multimodal images and clini-cal features can effectively predict the prognosis of patients with breast cancer.
Breast cancerMagnetic resonance imagingMultimodalNeoadjuvant chemotherapyPrognosisNo-mograph
NETL
NSTL
万方数据
