Evaluation of deformation parameters for deformable image registration-based ventilation imaging using an...

Area of Science:

• Medical Imaging
• Computational Anatomy
• Pulmonary Function Analysis

Background:

• Deformable image registration (DIR) is essential for analyzing dynamic physiological processes like lung ventilation.
• Evaluating the impact of DIR parameters on imaging accuracy is critical for clinical applications.
• NiftyReg is an open-source software package widely used for medical image registration.

Purpose of the Study:

• To assess the influence of various deformable image registration (DIR) parameters within the NiftyReg package on the accuracy of ventilation imaging.
• To identify optimal DIR parameter settings for reliable pulmonary ventilation visualization.

Main Methods:

• Two 3D-CT scans (exhalation and inhalation) of a ventilating phantom using xenon (Xe) contrast were acquired.
• Four DIR parameter sets were compared: one two-step and three four-step deformations.
• Spatial accuracy was evaluated using Target Registration Error (TRE) at 16 landmarks.
• Ventilation imaging accuracy was assessed by correlating Jacobian determinant (JD) metrics with Hounsfield unit (HU) changes.

Main Results:

• Four-step deformations achieved significantly lower mean TRE (1.47–1.56 mm) compared to two-step deformation (4.5 mm).
• Four-step deformations demonstrated stronger correlations (R = -0.71, -0.65, -0.61) between JD metrics and HU changes than two-step deformation (R = -0.40).
• Spatial accuracy was within acceptable limits across tested parameters.

Conclusions:

• The selection of DIR parameters significantly impacts the accuracy of DIR-based ventilation imaging.
• Adequate parameter settings for four-step NiftyReg DIR were identified for effective visualization of simulated pulmonary ventilation.
• While spatial accuracy may be tolerable, optimizing DIR parameters is key for robust ventilation imaging.