A vessel bifurcation landmark pair dataset for abdominal CT deformable image registration (DIR) validation

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Deformable image registration (DIR) is crucial for medical diagnostics and therapies.
• Clinical adoption of DIR is hindered by a lack of benchmark datasets for quality assurance.
• Abdominal CT registration presents unique challenges due to organ deformation and inconsistent imaging.

Purpose of the Study:

• Introduce a first-of-its-kind benchmark dataset for abdominal CT deformable image registration (DIR).
• Provide a dataset with a large number of highly accurate landmark pairs on matching blood vessel bifurcations.
• Facilitate the development and validation of advanced DIR algorithms for clinical applications.

Main Methods:

• Acquired abdominal CT image pairs from public repositories and institutional sources.
• Developed a workflow involving deep learning segmentation, manual landmark identification, and automated refinement.
• Generated 1895 landmark pairs with an average accuracy of 0.7 mm ± 1.2 mm.

Main Results:

• Created a unique dataset specifically for abdominal CT DIR validation.
• The dataset contains an average of 63 landmark pairs per patient case.
• Achieved high accuracy in landmark pair localization, crucial for rigorous algorithm testing.

Conclusions:

• The introduced dataset is the first of its kind for validating abdominal DIR.
• The dataset's size, accuracy, and distribution of landmarks enable precise DIR algorithm validation.
• This resource will significantly advance the development and clinical utility of DIR technologies.