MammoTracker: Mask-Guided Lesion Tracking in Temporal Mammograms

2025-06-30

AI Summary

To address the challenges of automatic cross-examination lesion matching in temporal mammograms and the limited progression analysis capability of existing computer-aided diagnosis (CAD) systems, this paper proposes a mask-guided lesion tracking framework. Methodologically, it adopts a coarse-to-fine three-stage strategy: global search for candidate regions, local search for precise localization, and score optimization for robust lesion association—integrating lesion mask priors, instance-level matching, and deep feature alignment. We introduce the largest temporal lesion tracking dataset to date, comprising over 20,000 lesion pairs. Experimental results demonstrate an 8% improvement over baseline methods in both overlap ratio (0.455) and accuracy (0.509), significantly enhancing CAD systems’ capacity for quantitative monitoring of breast cancer evolution and early diagnosis support.

Technology Category

• Computer Vision: Motion & Tracking
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Web Mining and Content Analysis: Topic discovery and tracking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Accurate lesion tracking in temporal mammograms is essential for monitoring breast cancer progression and facilitating early diagnosis. However, automated lesion correspondence across exams remains a challenges in computer-aided diagnosis (CAD) systems, limiting their effectiveness. We propose MammoTracker, a mask-guided lesion tracking framework that automates lesion localization across consecutively exams. Our approach follows a coarse-to-fine strategy incorporating three key modules: global search, local search, and score refinement. To support large-scale training and evaluation, we introduce a new dataset with curated prior-exam annotations for 730 mass and calcification cases from the public EMBED mammogram dataset, yielding over 20000 lesion pairs, making it the largest known resource for temporal lesion tracking in mammograms. Experimental results demonstrate that MammoTracker achieves 0.455 average overlap and 0.509 accuracy, surpassing baseline models by 8%, highlighting its potential to enhance CAD-based lesion progression analysis. Our dataset will be available at https://gitlab.oit.duke.edu/railabs/LoGroup/mammotracker.

Problem

• Automated lesion tracking in temporal mammograms for breast cancer monitoring
• Addressing challenges in lesion correspondence across exams for CAD systems
• Developing a large-scale dataset for training and evaluating lesion tracking models

Innovation

• Mask-guided lesion tracking framework
• Coarse-to-fine strategy with three modules
• New dataset with 20000 lesion pairs