Cross-attention guided multi-modal network for breast ultrasound diagnosis incorporating objective clinical semantics - Summary - MDSpire

Cross-attention guided multi-modal network for breast ultrasound diagnosis incorporating objective clinical semantics

  • By

  • Qin Sun

  • Xiaoman Wu

  • Ju Chen

  • Yuhang Zhang

  • Chao Zhou

  • Zheng Zhu

  • June 9, 2026

  • 0 min

Share

Objective:

To develop a multi-modal framework that integrates visual data with clinical descriptors for breast ultrasound diagnosis.

Key Findings:
  • CGA-Net achieved an Out-Of-Fold AUC of 0.905 and an overall accuracy of 0.857.
  • The model achieved a specificity of 0.831 and a sensitivity of 0.898.
  • A pre-trained version of CGA-Net reached a peak AUC of 0.915.
  • CGA-Net outperformed clinical-only (0.890) and image-only (0.795) baselines.
Interpretation:

Limitations:
  • The study relies on a specific dataset, which may limit generalizability.
  • Potential biases in the dataset could affect model performance.
Conclusion:

Original Source(s)

Cross-attention guided multi-modal network for breast ultrasound diagnosis incorporating objective clinical semantics

  • by Qin Sun, Xiaoman Wu, Ju Chen, Yuhang Zhang, Chao Zhou, Zheng Zhu

  • June 9, 2026

Related Content

Parvovirus B19 infection preceding the diagnosis of childhood myelodysplastic syndrome with low blasts: a case report

Tislelizumab combined with gemcitabine as first-line treatment in cisplatin-ineligible patients with locally advanced or metastatic urothelial carcinoma: a single center, single-arm phase 2 trial

TIFA: a signaling hub linking inflammation, innate immunity, and human disease