Quantitative assessment of pulmonary artery occlusion using lung dynamic perfusion CT

Abstract

Quantitative measurement of lung perfusion is a promising tool to evaluate lung pathophysiology as well as to assess disease severity and monitor treatment. However, this novel technique has not been adopted clinically due to various technical and physiological challenges; and it is still in the early developmental phase where the correlation between lung pathophysiology and perfusion maps is being explored. The purpose of this research work is to quantify the impact of pulmonary artery occlusion on lung perfusion indices using lung dynamic perfusion CT (DPCT). We performed Lung DPCT in ten anesthetized, mechanically ventilated juvenile pigs (18.6-20.2 kg) with a range of reversible pulmonary artery occlusions (0%, 40-59%, 60-79%, 80-99%, and 100%) created with a balloon catheter. For each arterial occlusion, DPCT data was analyzed using first-pass kinetics to derive blood flow (BF), blood volume (BV) and mean transit time (MTT) perfusion maps. Two radiologists qualitatively assessed perfusion maps for the presence or absence of perfusion defects. Perfusion maps were also analyzed quantitatively using a linear segmented mixed model to determine the thresholds of arterial occlusion associated with perfusion derangement. Inter-observer agreement was assessed using Kappa statistics. Correlation between arterial occlusion and perfusion indices was evaluated using the Spearman-rank correlation coefficient. Our results determined that perfusion defects were detected qualitatively in BF, BV and MTT perfusion maps for occlusions larger than 55%, 80% and 55% respectively. Inter-observer agreement was very good with Kappa scores > 0.92. Quantitative analysis of the perfusion maps determined the arterial occlusion threshold for perfusion defects was 50%, 76% and 44% for BF, BV and MTT respectively. Spearman-rank correlation coefficients between arterial occlusion and normalized perfusion values were strong (- 0.92, - 0.72, and 0.78 for BF, BV and MTT, respectively) and were statically significant (p < 0.01). These findings demonstrate that lung DPCT enables quantification and stratification of pulmonary artery occlusion into three categories: mild, moderate and severe. Severe (occlusion ≥ 80%) alters all perfusion indices; mild (occlusion < 55%) has no detectable effect. Moderate (occlusion 55-80%) impacts BF and MTT but BV is preserved.

Figure 1

An example of blood flow (BF, ml/100 ml/mn), blood volume (BV, ml/100 ml), and mean transit time (MTT, s) of perfusion maps generated for one of the pigs for all dynamic perfusion exams. A perfusion map color bar scale is also displayed at the right of each row to determine the corresponding perfusion values. Each column, which includes BF, BV and MTT maps, corresponds to one data set that was used to perform the qualitative perfusion defect detection using a randomized blinded approach. In this example, the two readers gave identical results for the presence/absence of perfusion defects. In (a,b), where occlusions were 0% and 40%, the readers indicated that no perfusion defect could be detected in any of the maps. In (d,e), where occlusions were 90% and 100%, the readers noted the presence of perfusion defect in all maps. In (c), where the occlusion was 70%, both readers noted the presence of defects in BF and MTT maps (highlighted with arrow) and absence of a perfusion defect in BV maps.

Figure 2

Right pulmonary artery occlusion with an inflatable balloon catheter and its impact on lung parenchyma time density curves (TDC). Figures (a,b) display axial and coronal slices with the balloon location highlighted using an arrow. The arterial occlusion for this experiment was 70% of the right pulmonary artery lumen. Figure (c), displays lung parenchyma enhancement at two regions of interest (ROI_O, distal to occluded artery; and ROI_N, in the normal lung parenchyma). The TDC distal to the balloon occlusion is characterized by a delay and a decrease in parenchyma enhancement.

Figure 3

Box-Whisker plots of normalized perfusion ratios of blood flow (BF, a), blood volume (BV, b), and mean transit time (MTT, c) from all 10 pigs and all occlusions. Measurements from each occlusion category (0, 40–59, 60–79, 80–99 and 100%) are grouped and compared to baseline (0% occlusion) using paired t-test. P-values are also presented to demonstrate the statistical significance of perfusion value change compared to baseline for each occlusion range.

Figure 4

Scatter plots of normalized perfusion ratios of blood flow (BF, a), blood volume (BV, b), and mean transit time (MTT, c) for all 10 pigs and for all occlusions displayed as markers (plus sign, multiple sign, open triangle, filled triangle, open circle, filled circle, open square, filled square, open diamond, filled diamond). Perfusion values from each pig is represented with individual markers. Results from piece-wise linear function fitting that contains two segments (Eq. 4) is also displayed as dashed lines. Fitted data demonstrates occlusion level onsets, A₀ = 50, 76, and 44%; as well as slopes of β = − 1.62, − 2.70, and 1.67 per %occlusion for BF, BV and MTT respectively.