Clinical Report: Confidence-Weighted Decision Fusion for Breast Ultrasound Lesions

Overview

This study presents a multi-scale information-bottleneck-guided classification framework for breast ultrasound (BUS) lesions. The proposed method utilizes a ResNet backbone and confidence-weighted decision-level fusion.

Background

Breast cancer is a leading cause of cancer-related mortality among women, and breast ultrasound (BUS) is a critical tool for early detection and treatment planning. However, the interpretation of BUS images is often hindered by operator variability and technical challenges.

Data Highlights

Experiments on breast ultrasound datasets were conducted to evaluate the proposed framework.

Key Findings

• The proposed framework utilizes a ResNet backbone with a feature pyramid network (FPN) for hierarchical multi-scale representation.
• An information bottleneck (IB) module is integrated at each FPN level.
• Auxiliary classifiers provide supervision at each scale.
• Confidence-weighted decision-level fusion aggregates predictions.

Clinical Implications

The proposed classification framework aims to enhance the accuracy of BUS lesion analysis.

Conclusion

The multi-scale information-bottleneck-guided classification framework shows potential in improving BUS lesion classification.

Related Resources & Content

1. Waks and Winer, Frontiers in Oncology, 2025 -- Breast cancer remains a leading cause of cancer-related mortality.
2. ACR Appropriateness Criteria® Female Breast Cancer Screening: 2025 Update, ScienceDirect, 2025 -- Guidelines for breast cancer screening.
3. ACR Publishes BI-RADS v2025 Manual to Advance Breast Imaging Standards, ACR, 2025 -- Updates in breast imaging standards.
4. Frontiers in Oncology — Multimodal feature fusion model for breast mass malignant risk stratification
5. Frontiers in Oncology — A synergistic framework integrating global context and structural features for breast ultrasound lesion detection
6. DIGITAL HEALTH — Breast lesion identification using feature fusion and multiresolution dual-tree complex wavelet transform
7. Frontiers in Oncology — Enhanced differentiation of breast lesions through integration of microvascular flow imaging and machine learning algorithms
8. Artificial intelligence in breast ultrasound: a systematic review of research advances
9. Screening for Breast Cancer - NCI
10. ACR Appropriateness Criteria® Female Breast Cancer Screening: 2025 Update - ScienceDirect
11. ACR Publishes BI-RADS v2025 Manual to Advance Breast Imaging Standards
12. Artificial Intelligence-Aided Detection of Breast Cancer Using Elastography: A Meta-Analysis of Diagnostic Test Accuracy
13. Artificial intelligence-enhanced handheld breast ultrasound for screening: A systematic review of diagnostic test accuracy - PubMed
14. Adjunct Automated Breast Ultrasound in Mammographic Screening: A Systematic Review and Meta-Analysis - PMC
15. ESR Essentials: artificial intelligence in breast imaging—practice recommendations by the European Society of Breast Imaging | European Radiology | Springer Nature Link
16. Artificial Intelligence-Enabled Medical Devices | FDA
17. FDA 510(k) clearance letter for Koios DS (K212616)