Semi-automatic construction of reference standards for evaluation of image registration

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Quantitative evaluation of image registration algorithms is challenging due to the absence of reference standards.
• Accurate evaluation is crucial for advancing medical imaging analysis and clinical applications.

Purpose of the Study:

• To develop an efficient, semi-automatic method for constructing detailed reference standard data for image registration.
• To enable robust quantitative evaluation of image registration algorithm performance.

Main Methods:

• A semi-automatic approach combining automatic landmark detection and manual observer input for correspondence.
• Utilizing a thin-plate-spline model to automatically match remaining landmarks based on observer-defined correspondences.
• Application to diverse datasets including thoracic CT and brain MR scans, and synthetic deformation data.

Main Results:

• Demonstrated accuracy of matched points using interobserver differences.
• Successfully generated reference standard data for multiple imaging modalities and datasets.
• Showcased the utility of the generated data in comparing multiple image registration algorithm results.

Conclusions:

• The proposed semi-automatic method efficiently creates high-quality reference standards for image registration.
• This approach addresses a critical gap in the quantitative evaluation of registration algorithms.
• Facilitates more reliable assessment and development of image registration techniques in medical imaging.