Interpretable Deep Transfer Learning for Breast Ultrasound Cancer Detection: A Multi-Dataset Study

2025-09-05

AI Summary

Early breast cancer screening demands AI-assisted diagnostic methods that are highly robust and interpretable. To address this, we propose an explainable deep transfer learning framework tailored for multi-center breast ultrasound images. Our method integrates multiple backbone architectures—including ResNet-18, EfficientNet-B0, and GoogLeNet—enhances deep features via ensemble fusion with traditional classifiers (SVM/KNN), and employs Grad-CAM to generate interpretable lesion heatmaps. In cross-dataset evaluation, the ResNet-18 variant achieves 99.7% accuracy and 100% sensitivity for malignant lesion detection, substantially outperforming single-dataset baselines. The framework demonstrates strong generalizability and robustness across heterogeneous clinical sites, while its built-in interpretability mechanisms significantly improve clinical trustworthiness. This work establishes a practical paradigm for intelligent ultrasound diagnosis that simultaneously ensures high diagnostic accuracy, broad generalizability, and model transparency.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Breast cancer remains a leading cause of cancer-related mortality among women worldwide. Ultrasound imaging, widely used due to its safety and cost-effectiveness, plays a key role in early detection, especially in patients with dense breast tissue. This paper presents a comprehensive study on the application of machine learning and deep learning techniques for breast cancer classification using ultrasound images. Using datasets such as BUSI, BUS-BRA, and BrEaST-Lesions USG, we evaluate classical machine learning models (SVM, KNN) and deep convolutional neural networks (ResNet-18, EfficientNet-B0, GoogLeNet). Experimental results show that ResNet-18 achieves the highest accuracy (99.7%) and perfect sensitivity for malignant lesions. Classical ML models, though outperformed by CNNs, achieve competitive performance when enhanced with deep feature extraction. Grad-CAM visualizations further improve model transparency by highlighting diagnostically relevant image regions. These findings support the integration of AI-based diagnostic tools into clinical workflows and demonstrate the feasibility of deploying high-performing, interpretable systems for ultrasound-based breast cancer detection.

Problem

• Breast cancer detection using ultrasound images
• Evaluating machine and deep learning classification techniques
• Enhancing model interpretability for clinical integration

Innovation

• Deep transfer learning with ResNet-18 architecture
• Grad-CAM visualizations for model interpretability
• Multi-dataset validation using BUSI and BUS-BRA