Applying an animal model to quantify the uncertainties of an image-based 4D-CT algorithm

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computational Imaging

Background:

• Four-dimensional computed tomography (4D-CT) is crucial for radiotherapy planning, enabling motion management.
• Accurate 4D-CT reconstruction relies on precise image registration and motion quantification.
• In vivo validation of 4D-CT algorithms is essential to assess their performance and assumptions.

Purpose of the Study:

• To quantify spatial displacement uncertainties of an image-based 4D-CT algorithm using an in vivo animal model.
• To test fundamental assumptions of the 4D-CT algorithm regarding respiratory phase matching and subvolume registration.
• To evaluate the accuracy of matching 3D subvolumes using single 2D overlapping slices.

Main Methods:

• An in vivo animal model (Landrace cross pigs) was used with a 64-slice CT scanner in axial cine mode.
• Breathing patterns were varied to simulate respiratory motion during image acquisition.
• Normalized cross-correlation (NCC) was employed to match CT images at the same respiratory phase, quantifying displacement uncertainties.

Main Results:

• Displacement uncertainty from NCC image matching ranged from 0.54 ± 0.10 mm to 0.32 ± 0.16 mm per match in the lung.
• Uncertainty propagated in quadrature, increasing with the number of NCC matches.
• The assumption that a 4.0 cm 3D subvolume can be matched by a single 2D overlapping slice was validated.

Conclusions:

• The study successfully developed an in vivo animal model for 4D-CT algorithm validation.
• Quantified uncertainties associated with NCC-based respiratory phase matching in 4D-CT.
• Validated a key assumption for efficient 3D subvolume matching in 4D-CT, supporting its promise for radiotherapy applications.