Image registration for DSA quality enhancement

Area of Science:

• Medical Imaging
• Computer Vision
• Image Processing

Background:

• Digital subtraction angiography (DSA) requires precise image registration for effective enhancement.
• Existing similarity measures may not be optimal for the specific challenges in DSA, such as motion and distortions.

Purpose of the Study:

• To present a generalized framework for histogram-based similarity measures.
• To apply this framework to the image-enhancement task in DSA.
• To identify and evaluate a robust and computationally efficient similarity measure for DSA image registration.

Main Methods:

• Developed a generalized framework for histogram-based similarity measures using differentiable, strictly convex weighting functions.
• Identified the energy similarity measure as a suitable function for registering mask and contrast images in DSA.
• Investigated the robustness of the energy measure against geometrical distortions (rotation, scaling) and factors like histogram binning and motion.

Main Results:

• The energy similarity measure is computationally efficient for DSA image registration.
• The energy measure shows robustness to geometrical image distortions.
• The automated registration procedure using the energy measure achieved comparable success to manual correction for head DSA.

Conclusions:

• The proposed generalized framework and the energy similarity measure offer an effective solution for DSA image registration and enhancement.
• The energy measure provides a robust and computationally efficient approach for clinical applications of DSA.