Letter to the Editor

Interpretability and external validation remain key barriers to clinical integration of multimodal deep learning for breast cancer prognosis