MGRegBench: A Novel Benchmark Dataset with Anatomical Landmarks for Mammography Image Registration

TL;DR

MGRegBench introduces a large public 2D mammography registration benchmark with 100 landmark-annotated image pairs and a substantial training set, drawn from INBreast, KAU-BCMD, and RSNA. It provides a standardized evaluation framework, comprehensive metrics, and baseline results across classical, implicit, and deep learning methods, highlighting that an affine pre-alignment followed by a deformable MRN model yields the best overall performance. The study demonstrates that while deep learning methods offer strong intensity-based and segmentation metrics, classical methods remain competitive for anatomical accuracy, and hybrid approaches can offer robust, clinically relevant performance. By releasing data, protocols, and implementations, MGRegBench aims to catalyze reproducible research and accelerate development of robust longitudinal mammography analysis tools.

Abstract

Robust mammography registration is essential for clinical applications like tracking disease progression and monitoring longitudinal changes in breast tissue. However, progress has been limited by the absence of public datasets and standardized benchmarks. Existing studies are often not directly comparable, as they use private data and inconsistent evaluation frameworks. To address this, we present MGRegBench, a public benchmark dataset for mammogram registration. It comprises over 5,000 image pairs, with 100 containing manual anatomical landmarks and segmentation masks for rigorous evaluation. This makes MGRegBench one of the largest public 2D registration datasets with manual annotations. Using this resource, we benchmarked diverse registration methods including classical (ANTs), learning-based (VoxelMorph, TransMorph), implicit neural representation (IDIR), a classic mammography-specific approach, and a recent state-of-the-art deep learning method MammoRegNet. The implementations were adapted to this modality from the authors' implementations or re-implemented from scratch. Our contributions are: (1) the first public dataset of this scale with manual landmarks and masks for mammography registration; (2) the first like-for-like comparison of diverse methods on this modality; and (3) an extensive analysis of deep learning-based registration. We publicly release our code and data to establish a foundational resource for fair comparisons and catalyze future research. The source code and data are at https://github.com/KourtKardash/MGRegBench.

Figures (6)

• Figure 1: Example of a full-fledged paired study from INBreast dataset made in 2009 and in 2010 year correspondingly. This case is particularly illustrative due to the presence of metallic surgical clips in the right breast, indicating a prior surgical intervention.
• Figure 2: Example of a registration-ready image pair from each source dataset included in MGRegBench: (left) INBreast, (center) KAU-BCMD, and (right) RSNA. Each pair shows the same breast (left or right) in the same projection (CC or MLO) acquired during separate screening exams.
• Figure 3: Overlaid histograms of patient age and breast density distributions in the training (blue, left y-axis) and evaluation (olive, right y-axis) sets. The training histograms are plotted in the background, while the evaluation histograms are overlaid in the foreground for direct visual comparison. The x-axis represents density or age category, and the two y-axes show the respective case counts.
• Figure 4: Examples of anatomical landmarks used for registration evaluation in MGRegBench. Each panel illustrates a distinct landmark type manually annotated by expert radiologists: microcalcifications, bends in ducts or blood vessels, vessel or duct intersections, forks, visible masses, and dark/bright blob contours.
• Figure 5: Example of landmark annotation performed by two radiologists. Blue markers indicate landmarks placed during the initial annotation phase; yellow markers show the refined annotations from the second (validation) phase.