Reproducibility of perfusion blood volume and ventilation imaging for lung function assessment

Objective.To assess the reproducibility of dual-energy CT (DECT) derived perfusion blood volume (PBV) and four-dimensional CT (4DCT) derived lung ventilation images.Approach.We enrolled patients on an IRB-approved clinical trial of functional lung imaging. All had locally advanced non-small cell lung cancer (NSCLC) treated with definitive, conventionally fractionated concurrent chemoradiation radiotherapy (RT) with or without consolidation immunotherapy. Patients completed imaging at three timepoints: (1) RT simulation, (2) second-week of RT, and (3) six-months post-RT. Free-breathing 4DCT and contrast-enhanced DECT were acquired at each timepoint. To evaluate reproducibility of PBV and ventilation maps, voxel intensities within the lung parenchyma from exclusively the first two timepoints were binned according to relative image intensity. Binary masks were generated at 20%, 40%, 60%, and 80% thresholds. Dice-similarity-coefficients (DSC) and structural-similarity-index-measure (SSIM) were calculated between the corresponding masks to assess the robustness and reproducibility. Scatter plots were generated to visualize voxel-wise intensity agreement.Main Results.Reproducibility of ventilation and perfusion images was evaluated in 28 patients between baseline and 2 week follow-up. For perfusion, the 20%-100% threshold showed the highest DSC (Scanner A = 0.91 ± 0.06; Scanner B = 0.81 ± 0.08), with progressive decreases at higher thresholds (Scanner A: 40%-100%(0.81 ± 0.08), 60%-100%(0.65 ± 0.07), and 80%-100%(0.48 ± 0.07); Scanner B: 40%-100%(0.73 ± 0.13), 60%-100%(0.58 ± 0.08), and 80%-100%(0.39 ± 0.07), respectively). Ventilation images showed a similar trend, with DSC of 20%-100%(0.83 ± 0.05), 40%-100%(0.71 ± 0.07), 60%-100%(0.59 ± 0.10), and 80%-100%(0.46 ± 0.11). For perfusion, mean SSIM for Scanner A decreased from 20%-100%(0.72 ± 0.10) to 80%-100%(0.66 ± 0.14); Scanner B from 20%-100%(0.55 ± 0.10) to 80%-100%(0.54 ± 0.08). For ventilation, the SSIM had a similar trend and declined from 20%-100%(0.73 ± 0.11) to 80%-100%(0.68 ± 0.15).Significance.DECT-derived perfusion showed superior reproducibility to 4DCT ventilation between timepoints 1 and 2, with consistently higher DSC and SSIM across thresholds. Despite systematic perfusion increases early in treatment, its spatial stability remained stronger, supporting DECT perfusion as a more robust biomarker for longitudinal lung function assessment and for informing adaptive, functionally-guided RT in NSCLC.