Introducing a novel sub-millimeter lung CT image registration error quantitation tool.

Area of Science:

• Medical imaging
• Image processing
• Radiology

Background:

• Lung CT image registration is vital for functional analyses like ventilation.
• High sensitivity of functional analyses necessitates precise sub-millimeter error detection.
• Existing tools lack the required sensitivity for critical registration error measurement.

Purpose of the Study:

• To introduce lambda, an image registration error scoring tool.
• To quantify registration errors in steep image gradient regions under realistic noise.
• To achieve high spatial sensitivity for sub-millimeter error detection.

Main Methods:

• Lambda normalizes image scales and calculates minimum Euclidean distances between voxels.
• Simulated blood vessels with varying diameters and noise levels were used for testing.
• Rigid translational and rotational deformations assessed lambda's ability to track shifts.
• Restricted lambda (theta) was used as a proxy for image gradient, with 95th percentile determining spatial sensitivity.

Main Results:

• Lambda's spatial component accurately tracked simulated vessel shifts.
• The spatial sensitivity limit of lambda was determined to be less than 0.2 mm.
• Increasing noise and smoothing kernel parameters affected lambda's sensitivity.
• Clinical CT scans showed lambda qualitatively matching intensity differences and highlighting error regions.

Conclusions:

• Lambda successfully detected sub-millimeter positioning errors in simulated lung CT scans with noise.
• Noise magnitude and smoothing kernel choice inversely impacted lambda's sensitivity.
• Further evaluation of lambda for detecting sub-voxel registration errors is warranted.