多模态深度学习融合模型：开启HER2阳性与三阴性乳腺癌新辅助治疗疗效精准预测新篇章
Ranking
No observed public metrics; popularity remains neutral/archived.
Merged summary
TL;DR - A multicenter proof-of-concept study combined pretreatment breast imaging and clinical features to predict pathological complete response after neoadjuvant therapy in HER2-positive and triple-negative breast cancer. The fusion model could support personalized treatment decisions, but requires larger prospective validation.
- The study included 359 patients from two institutions and fused six mammography/mpMRI sequences using DenseNet169-CBAM.
- The clinical-imaging fusion model performed best, with AUCs of 0.992, 0.877, and 0.768 in the training, validation, and external test sets.
- Minimum and mean ADC values and tumor margin were independent predictors; SHAP identified minimum ADC as the most influential feature.
- Limited sample size, subtype imbalance, and reliance on pretreatment imaging constrain generalizability.
Sources (1)
多模态深度学习融合模型：开启HER2阳性与三阴性乳腺癌新辅助治疗疗效精准预测新篇章
TL;DR - A multicenter proof-of-concept study combined pretreatment breast imaging and clinical features to predict pathological complete response after neoadjuvant therapy in HER2-positive and triple-negative breast cancer. The fusion model could support personalized treatment decisions, but requires larger prospective validation.
- The study included 359 patients from two institutions and fused six mammography/mpMRI sequences using DenseNet169-CBAM.
- The clinical-imaging fusion model performed best, with AUCs of 0.992, 0.877, and 0.768 in the training, validation, and external test sets.
- Minimum and mean ADC values and tumor margin were independent predictors; SHAP identified minimum ADC as the most influential feature.
- Limited sample size, subtype imbalance, and reliance on pretreatment imaging constrain generalizability.