Objective:

To develop and validate a nomogram integrating multimodal imaging and clinical features for the preoperative prediction of HER2 expression status in patients with invasive breast cancer.

Approach:

• Study Design: Retrospective study involving 204 patients with pathologically confirmed breast cancer who underwent preoperative DCE-MRI and mammography.
• Feature Selection: Optimal radiomics features were selected using Pearson correlation coefficient combined with recursive feature elimination.
• Model Development: Three binary classification tasks were constructed to differentiate HER2 expression states, with models based on MG, DCE-MRI, and their combination.
• Performance Evaluation: The performance of the models was assessed using area under the curve (AUC) values in both training and validation sets.

Key Findings:

• The combined radiomics model achieved AUC values of 0.742 (95% CI: 0.624–0.860) and 0.823 (95% CI: 0.749–0.897) for Model_1 and Model_2 in the training set.
• Validation set AUC values were 0.718 (95% CI: 0.594–0.842) and 0.778 (95% CI: 0.696–0.861) for Model_1 and Model_2, respectively.
• The DCE-MRI model showed optimal performance for Model_3 with an AUC of 0.831 (95% CI: 0.747–0.916) in the training set and 0.745 (95% CI: 0.640–0.850) in the validation set.

Interpretation:

The multimodal radiomics-based nomogram provides a non-invasive quantitative tool for preoperative evaluation of HER2 expression status in breast cancer.

Limitations:

• The study was retrospective and conducted at a single institution.
• No nomogram was developed for Model_3 due to the absence of significant clinicopathological features.

Conclusion:

The integration of imaging features and clinical factors in a multimodal radiomics-based nomogram shows potential for assessing HER2 expression status.