Automated 3D Segmentation of the Aorta and Pulmonary Artery on Non-Contrast-Enhanced Chest Computed Tomography Images in Lung Cancer Patients

DRU-Net: Pulmonary Artery Segmentation via Dense Residual U-Network with Hybrid Loss Function

The structure and topology of the pulmonary arteries is crucial to understand, plan, and conduct medical treatment in the thorax area. Due to the complex anatomy of the pulmonary vessels, it is not easy to distinguish between the arteries and veins. The pulmonary arteries have a complex structure with an irregular shape and adjacent tissues, which makes automatic segmentation a challenging task. A deep neural network is required to segment the topological structure of the pulmonary artery. Therefore, in this study, a Dense Residual U-Net with a hybrid loss function is proposed. The network is trained on augmented Computed Tomography volumes to improve the performance of the network and prevent overfitting. Moreover, the hybrid loss function is implemented to improve the performance of the network. The results show an improvement in the Dice and HD95 scores over state-of-the-art techniques. The average scores achieved for the Dice and HD95 scores are 0.8775 and 4.2624 mm, respectively. The proposed method will support physicians in the challenging task of preoperative planning of thoracic surgery, where the correct assessment of the arteries is crucial.

Automated 3D segmentation of the aorta and pulmonary artery for predicting outcomes after thoracoscopic lobectomy in lung cancer patients

Background

Preoperative two-dimensional manual measurement of pulmonary artery diameter in a single-cut axial view computed tomography (CT) image is a commonly used non-invasive prediction method for pulmonary hypertension. However, the accuracy may be unreliable. Thus, this study aimed to evaluate the correlation of short-term surgical outcomes and pulmonary artery/aorta (PA/Ao) diameter ratio measured by automated three-dimensional (3D) segmentation in lung cancer patients who underwent thoracoscopic lobectomy.

Materials and methods

We included 383 consecutive lung cancer patients with thin-slice CT images who underwent lobectomy at a single institute between January 1, 2011 and December 31, 2019. Automated 3D segmentation models were used for 3D vascular reconstruction and measurement of the average diameters of Ao and PA. Propensity-score matching incorporating age, Charlson comorbidity index, and lobectomy performed by uniportal VATS was used to compare clinical outcomes in patients with PA/Ao ratio ≥1 and those <1.

Results

Our segmentation method measured 29 (7.57%) patients with a PA/Ao ratio ≥1. After propensity-score matching, a higher overall postoperative complication classified by the Clavien–Dindo classification (p = 0.016) were noted in patients with 3D PA/Ao diameter ratio ≥1 than those of <1. By multivariate logistic regression, patients with a 3D PA/Ao ratio ≥ 1 (p = 0.013) and tumor diameter > 3 cm (p = 0.002) both significantly predict the incidence of postoperative complications.

Conclusions

Pulmonary artery/aorta diameter ratio ≥ 1 measured by automated 3D segmentation may predict postoperative complications in lung cancer patients who underwent lobectomy.