This study was conducted in accordance with the Declaration of Helsinki. In agreement with our local IRB, due to the data being registered anonymously and the study’s retrospective design, formal ethical approval by the IRB was not required.

Patients with moderate-severe PE, in evaluation for endoscopic LVR, with pulmonary angio-DECT imaging between February 2018 and March 2022, who underwent planar LPS within 15 days of the DECT at our Institution (Azienda Ospedaliero Universitaria delle Marche) were included. Patients with imaging of inadequate quality (i.e., pulmonary artery enhancement < 200 HU, motion artifacts, underlying pulmonary diseases such as consolidations or pulmonary embolism), or with incomplete imaging data (DECT or LPS data not available) were not included.

CT examinations were performed using 3rd generation Dual Source DECT (Somatom Force, Siemens Healthineers). The basal acquisition was performed in deep inspiration and reconstructed with a medium and sharp kernel (120 kV, modulated mA, model-based iterative reconstruction, ADMIRE, strength 3, kernels Br40 and Bl64) with a slice thickness and spacing of 1 mm/0.7 mm [34, 35]. The study was completed with a DECT acquisition (90/150Sn, modulated mA, ADMIRE 3, reconstruction kernels Br40, BL64, and Qr40, thickness/spacing: 1/0.7 mm) after the administration of intravenous contrast material (Iopamidol 370 mgI/mL, Iopamiro 370, Bracco; 1 mL/kg, 3,5 – 5 ml/s); the acquisition time was optimized with bolus tracking technique (Threshold: 100 HU in pulmonary trunk; acquisition delay: 6 s).

Visual quantification of emphysema was performed with Goddard VS on HRCT basal images by two readers in consensus and blinded to LS (AA, AB, with 15 and 10 years of experience in thoracic imaging), after the quality control of the DECT study (i.e., adequate enhancement of the pulmonary artery) and exclusion of concomitant pulmonary diseases [20]. Both lungs were evaluated at three levels: upper, middle, and lower. The apical zone was examined at an axial scan passing 1 cm above the upper margin of the aortic arch; the middle lung was examined at an axial scan passing 1 cm below the tracheal carina; the basal lung was examined 3 cm above the diaphragmatic dome. A percentage score of emphysematous parenchyma was assigned at these zones, bilaterally. In particular, no evidence of emphysema: score 0; emphysematous areas less than 25%: score 1; between 26 and 50%: score 2; between 51 and 75%: score 3; greater than 75%: score 4. These scores were summed, and the total value was the severity score of PE. Conventionally, the three regions of the lung considered for the visual analysis, and for the further segmentation (see below), were labeled as left (L1, L2, L3) and right (R1, R2, R3) in cranio-caudal direction.

Basal images reconstructed with medium kernel (Br40) were processed with semiautomatic software for densitometric analysis (Pulmo3D on Syngo.via, Siemens Healthineers) by the same radiologists blinded to LS [34, 35]. The software quantifies the percentage of voxels with attenuation values lower than the threshold of -950 HU for the entire pulmonary parenchyma, the right and left lungs separately and the lower, middle and upper zones of equal volume (Volume Based, VB) of each lung separately (R1-3; L1-3). The Mean Lung Densities (MLD) were also recorded for each patient.

The two radiologists in consensus and blinded to LS processed the post-contrast DECT datasets with dedicated three-material decomposition on DE Lung Analysis (Syngo.via, Siemens Healthineers) to calculate the lung PBV. Iodine distribution and PBV maps were rendered on axial, sagittal, and coronal planes and in 3D Volume Rendering (VR). The measurements were performed after normalization of iodine distribution on the Pulmonary artery enhancement: a standardized region of interest (ROI) sized 0.5 cm2 was placed in the pulmonary trunk on axial images. Lung partitioning was performed automatically using the Distance Based (DB) mode. The lungs were divided by two horizontal lines into three fields of equivalent cranio-caudal length: upper, middle, and lower (R1-3; L1-3). DB is a mode of lung segmentation more similar to that used in scintigraphy than the technique of VB segmentation used in densitometric analysis.

Perfusion Scintigraphy examinations were performed with γ-chamber Infinia (GE Healthcare). Eight standard planar projections (anterior, posterior, right lateral, left lateral, right posterior oblique, left posterior oblique, right anterior oblique, left anterior oblique) were acquired after intravenous administration of albumin macroaggregates labeled with 99mTechnetium (99mTc-MAA) [4]. The post-processing and image revision was performed with dedicated software by two nuclear medicine physicians in consensus and blinded to the DECT datasets (LB, AP, 30 and 10 years of experience respectively): each anterior and posterior digital planar scintigraphy projection of each lung was manually demarcated within a rectangular ROI which is semi-manually subdivided into three regions (upper, middle, lower) of approximately equivalent areas by two horizontal lines (R1-3; L1-3). The anteroposterior geometric mean of the detected counts was automatically calculated for all areas corresponding to anterior and posterior projections of each lung. The estimated lung perfusion was obtained in absolute k-count value for each lung field examined [26].