Mammogram registration: a phantom-based evaluation of compressed breast thickness variation effects

Area of Science:

• Medical Imaging
• Radiology
• Computer Vision

Background:

• Comparing mammograms over time is challenging due to factors altering image appearance.
• Mammogram registration aims to mitigate these changes, highlighting genuine breast tissue alterations.
• Evaluating registration methods is difficult due to the lack of ground truth for breast deformations in clinical images.

Purpose of the Study:

• To propose a systematic approach for evaluating the sensitivity of mammogram registration methods.
• To assess registration techniques using synthetic breast images with known deformations.
• To evaluate a nonrigid mammogram registration technique using simulated physical compression.

Main Methods:

• Utilized synthetic breast images with known, simulated physical compression variations.
• Employed a previously described nonrigid mammogram registration technique.
• Measured registration performance via average displacement error at defined evaluation points.

Main Results:

• Achieved an average displacement error of 1.6 mm for mammograms with 1-3 cm compression differences.
• Demonstrated the effectiveness of thickness compensation and preferred image ordering.
• Validated the proposed methodology for assessing registration sensitivity to mammogram variations.

Conclusions:

• The proposed systematic approach effectively evaluates mammogram registration sensitivity using synthetic data.
• The methodology can be extended to analyze other sources of mammogram differences.
• This approach can be adapted for registering multimodality breast data.