Prediction of the short-term efficacy of neoadjuvant chemotherapy for triple-negative breast cancer by ultrasound combined with MRI deep learning radiomics nomogram
10.3969/j.issn.1002-1671.2025.11.007
- VernacularTitle:超声联合MRI深度学习影像组学诺谟图预测三阴性乳腺癌新辅助化疗近期疗效
- Author:
Chen CHENG
1
;
Hongyan ZHAO
;
Yan WANG
;
Hong'e LI
;
Yan GU
;
Wenrong WANG
;
Feng XU
;
Jin'e ZHAO
Author Information
1. 连云港市中医院超声科,江苏 连云港 222004;南京医科大学康达学院临床医学院,江苏 连云港 222000
- Publication Type:Journal Article
- Keywords:
ultrasound;
magnetic resonance imaging;
deep learning radiomics nomogram;
triple-negative breast cancer
- From:
Journal of Practical Radiology
2025;41(11):1791-1796
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the predictive value of ultra-sound(US)combined with multi-sequence MRI in a clinical-deep learning radiomics nomogram(DLRN)for the short-term efficacy of neoadjuvant chemotherapy(NAC)in patients with triple-negative breast cancer(TNBC).Methods A total of 122 TNBC patients from five hospitals were retrospectively analyzed,and divided into training group(72 cases)and validation group(50 cases).The clinical and pathological data,NAC regimens,and imaging data were collected.The lesions and its surrounding 10-unit voxels from US and MRI images were retained as the region of interest(ROI).Pyradiomics software and a ResNet152 convolutional neural network(CNN)framework were used to extract radiomics and deep learning features to construct a clinical-DLRN with outcome dimension fusion.The receiver operating characteristic(ROC)curve and calibration curve were plotted,and five-fold cross-validation decision curve were used to verify the model's clinical effec-tiveness.Results The clinical-DLRN constructed by US combined with multi-sequence MRI showed that area under the curve(AUC)was 0.967[95％confidence interval(CI)0.782-0.967]and accuracy(ACC)was 0.900,respectively.The five-fold cross-validation decision curve showed good generalization,with the highest clinical net benefit between risk thresholds of 0.72 and 0.96.Conclusion The clinical-DLRN integrating US and multi-sequence MRI has the best efficacy in predicting the short-term efficacy of NAC in TNBC patients,which offering potential guidance for personalized TNBC treatment.
