The availability of dual-energy computed tomography (DECT) has allowed for more detailed study of lung perfusion. CT based studies have demonstrated pruning in thromboembolic lung diseases as well as perfusion variability but the relationship between pruning and perfusion variability is not well characterized. The purpose of this analysis was to study the variability in perfusion using DECT in a cohort of patients with and without thromboembolic disease and to compare this to pruning in the same subjects.
All DECTs with complete original information that were performed on the Siemens platform were retrospectively identified. All DECT (were reconstructed at high resolution (0.8mm thickness slices) and the pulmonary vasculature was reconstructed from the same images, and the arterial and venous vasculature were labeled using a convolutional neural network, validated previously by manually labeling. Measures of pulmonary blood volume were then extracted, with exact sizing provided by a deep learning based neural network. Venous and arterial BV5 and TBV (vessels <5mm2 in cross-section and total pulmonary blood vessel volume) were measured. The BV5/TBV ratio, a marker of pruning, was calculated for the arterial and venous systems
There were 87 DECTs included in the final cohort including 19 subjects with CTEPH, 11 subjects with CTED, 33 patients with acute PE and 24 subjects with no pulmonary thromboembolic disease by imaging. Upon analysis of the entire cohort, increased aBV5/TBV correlated with a decreased standard deviation of the normalized iodine perfusion map voxel values (R = 0.54). A similar relationship was also seen in the venous BV5/TBV (R=0.62). For results tables, see full pdf
In summary, patients with more distal pruning tended to have greater variability in their perfusion imaging than those with mild pruning. This suggests that pulmonary vascular remodeling on a macroscopic scale is associated changes in lung perfusions in thromboembolic disease.