A multimodal fusion method, system, medium, and product for breast cancer pathological diagnosis
By combining the Transformer model and multi-task learning technology, cross-modal deep fusion of ultrasound images and pathological text is achieved, which solves the problem of poor multimodal data fusion in existing breast cancer diagnosis and improves the accuracy and robustness of breast cancer pathological diagnosis.
CN122314359APending Publication Date: 2026-06-30ARTIFICIAL INTELLIGENCE RES INST OF HEFEI COMPREHENSIVE NAT SCI CENT (ANHUI ARTIFICIAL INTELLIGENCE LAB)
View PDF 0 Cites 0 Cited by
Patent Information
- Application Number
- CN202610632799.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-09
- Publication Date
- 2026-06-30
Smart Images

Figure CN122314359A_ABST
Abstract
This invention discloses a multimodal fusion method, system, medium, and product for breast cancer pathological diagnosis, relating to the interdisciplinary field of artificial intelligence and medical diagnostics. The method inputs ultrasound images and pathological texts of breast cancer patients into a multimodal fusion model to obtain multimodal fusion features and lesion segmentation masks. The training process of the multimodal fusion model is as follows: A training set is constructed after preprocessing the acquired ultrasound images and pathological texts; text-encoded features and image-encoded features are extracted from the training set; cross-modal attention fusion is performed on the text-encoded features and image-encoded features to obtain multimodal fusion features; based on the multimodal fusion features, the predicted probability of benign or malignant breast cancer is output through a pathological classification branch, and the lesion segmentation mask is output through a lesion segmentation branch. This method achieves simultaneous optimization of benign / malignant breast cancer classification and lesion region segmentation.
Need to check novelty before this filing date? Find Prior Art
