ABSTRACT This study aimed to develop a noninvasive nomogram that integrates deep learning‐pathomics, radiomics, and immunoscore to predict lymph node metastasis (LNM) in breast cancer. Pathological features from 1133 TCGA‐BRCA slides were extracted via ResNet50 and Lasso. Radiomics features from 137 MRI images (TCIA) were analyzed using pyradiomics. Immunoscore was calculated via ESTIMATE. A nomogram was constructed and validated with 10‐fold cross‐validation. The pathomics model achieved an AUC of 0.65 (95% CI: 0.61–0.68), sensitivity 0.62, specificity 0.67; radiomics 0.61 (95% CI: 0.50–0.72), sensitivity 0.59, specificity 0.63; and the combined nomogram 0.69 (95% CI: 0.59–0.79), sensitivity 0.66, specificity 0.71. Radiomics score was the strongest predictor. The nomogram provides a reliable noninvasive tool for predicting lymph node involvement, potentially reducing unnecessary biopsies.
Xu et al. (Thu,) studied this question.