Abstract
Given a deformation field by an image registration algorithm, significant differences between the SAJ, SACJ, and SAI measures were found at a regional level compared to the Xe-CT sV in four sheep that were studied. The SACJ introduced here, provides better correlations with Xe-CT based sV than the SAJ and SAI measures, thus providing an improved surrogate for regional ventilation.
Introduction
Understanding regional lung ventilation—the specific distribution of air within the lungs—is critical for diagnosing and monitoring pulmonary diseases such as COPD, asthma, and lung cancer. While global pulmonary function tests provide an overall assessment of lung health, they cannot identify localized areas of dysfunction.
Currently, Xenon-enhanced CT (Xe-CT) is considered the "gold standard" for regional ventilation mapping. However, Xe-CT presents significant logistical challenges, including the use of expensive radioactive gas, specialized equipment, and limited spatial coverage. A promising alternative involves using image registration on 4DCT (dynamic CT) scans. By taking images at different respiratory phases (e.g., end-inspiration and end-expiration) and applying non-rigid registration, researchers can estimate local air movement. However, several mathematical models exist to calculate ventilation from these deformation fields, and their relative accuracy has remained unclear.
Methodology
This study compared three registration-based measures of ventilation against the Xe-CT gold standard:
- SAJ (Specific Air volume change by Jacobian): A model assuming the lung consists entirely of air, using the Jacobian determinant of the deformation field to estimate expansion or contraction.
- SAI (Specific Air volume change by Intensity change): A method based on Hounsfield Unit (HU) changes, assuming that while air volume fluctuates, the mass of lung tissue within a voxel remains constant.
- SACJ (Specific Air volume change by Corrected Jacobian): A novel measure proposed by the authors that combines geometric warping (Jacobian) with density information (Intensity) to account for simultaneous changes in both air and tissue volumes.
The researchers utilized a "Tissue Volume and Vesselness Preserving" (TVP) non-rigid registration algorithm to ensure anatomical accuracy, using pulmonary blood vessels as internal landmarks. The study was conducted using data from four adult sheep, involving both 4DCT scans across eight respiratory phases and a series of 45 Xe-CT scans to track Xenon wash-in.
Key Results and Insights
The comparison revealed that the proposed SACJ method significantly outperformed existing models at a regional level.
- Regional Accuracy: When evaluated in 20mm cubes, SACJ showed the strongest correlation with the Xe-CT gold standard ($r^2 = 0.82$). The air-only SAJ followed ($r^2 = 0.75$), while the intensity-based SAI performed the worst ($r^2 = 0.58$).
- Scale Dependency: Interestingly, at a "global" scale (large lung slabs), all three methods performed similarly well, with $r^2$ values ranging from 0.87 to 0.92. This suggests that the flaws in simpler models only become apparent when high-resolution regional mapping is required.
- Mathematical Relationship: The authors demonstrated that the widely used SAI method is essentially a simplified version of the SACJ method. SAI's accuracy depends on the assumption that tissue volume never changes—an assumption that is physically incomplete in a dynamic lung.
Conclusion
By incorporating both geometric deformation and density changes, the SACJ method provides a more physically realistic and robust surrogate for regional air exchange than previous models. While the study was limited by a small sample size and the inherent noise of Xe-CT, it establishes a clear path forward for high-resolution ventilation mapping using standard CT imaging. Future work will focus on validating these measures in diseased lung models where tissue volume changes may be more pronounced.
Summary updated