Deep Learning-Based Breast Cancer Detection in Mammography: A Multi-Center Validation Study in Thai Population

2025-05-29

AI Summary

Addressing the urgent need for effective breast cancer screening in the Thai population, this study develops a clinically deployable, multi-center generalizable AI detection system. Method: We propose an enhanced EfficientNetV2 architecture incorporating a hybrid channel-spatial attention mechanism, integrated with lesion localization heatmap generation, multi-center data distribution adaptation during training, and a robustness evaluation framework for LLF/NLF localization. Contribution/Results: This work represents the first systematic validation of AI generalizability and radiologist-AI collaborative performance in real-world, multi-center Thai clinical settings, establishing a novel clinical concordance analysis paradigm. Experiments demonstrate an AUROC of 0.89–0.96, lesion localization LLF > 0.83, radiologist-AI concordance rates of 83.5% (classification) and 84.0% (localization), clinical acceptance > 89%, and usability score of 74.17 (source hospital), confirming high clinical applicability.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Efficient ML / Green AI
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

This study presents a deep learning system for breast cancer detection in mammography, developed using a modified EfficientNetV2 architecture with enhanced attention mechanisms. The model was trained on mammograms from a major Thai medical center and validated on three distinct datasets: an in-domain test set (9,421 cases), a biopsy-confirmed set (883 cases), and an out-of-domain generalizability set (761 cases) collected from two different hospitals. For cancer detection, the model achieved AUROCs of 0.89, 0.96, and 0.94 on the respective datasets. The system's lesion localization capability, evaluated using metrics including Lesion Localization Fraction (LLF) and Non-Lesion Localization Fraction (NLF), demonstrated robust performance in identifying suspicious regions. Clinical validation through concordance tests showed strong agreement with radiologists: 83.5% classification and 84.0% localization concordance for biopsy-confirmed cases, and 78.1% classification and 79.6% localization concordance for out-of-domain cases. Expert radiologists' acceptance rate also averaged 96.7% for biopsy-confirmed cases, and 89.3% for out-of-domain cases. The system achieved a System Usability Scale score of 74.17 for source hospital, and 69.20 for validation hospitals, indicating good clinical acceptance. These results demonstrate the model's effectiveness in assisting mammogram interpretation, with the potential to enhance breast cancer screening workflows in clinical practice.

Problem

• Develop deep learning for breast cancer detection in mammograms
• Validate model across diverse Thai population datasets
• Assess clinical concordance and radiologist acceptance rates

Innovation

• Modified EfficientNetV2 with enhanced attention
• Multi-center validation on diverse datasets
• High concordance with radiologists' assessments