To explore the value of a multimodal deep learning model integrating MRI, ultrasound, digital pathology and clinical information in predicting pCR and long-term prognosis.