Publication:  Cheng, H., Zheng, L., Yan, Z., Zhang, H., Meng, B., & Xu, X. (2024, December). Fusion of Machine Learning and Deep Neural Networks for Pulmonary Arteries and Veins Segmentation in Lung Cancer Surgery Planning. In International Conference on Pattern Recognition (pp. 422-438). Cham: Springer Nature Switzerland.
 
Scanner: Siemens’ SOMATOM Definition Flash
 
Number of Patients: N=106
 
Hospital: Guangdong Provincial Peoples’ Hospital
 
Description: Our dataset comprises 95 3D CT scans obtained from Siemens’ SOMATOM Definition Flash machine. The patients’ age ranges from 29 to 82 years, with an average of 58.4 years. The images have a size of 512 × 512×(280−370) voxels, with a typical voxel size of 0.25×0.25×0.5mm3. The annotations encompass intrapulmonary and extrapulmonary arteries and veins. All the annotations are performed by lung cancer expert surgeons, investing 2-3.5 hours for each image. It’s important to note that all patients in the dataset underwent lung cancer surgery, and these annotations have been effectively employed in clinical practice to support surgeons in planning such surgeries. For the sake of labeling efficiency, only vessels in the left or right lung with nodes were labeled. Consequently, most images have annotations for only half of the lung, while a few have labels for both sections. In our experiments, we divided the dataset into two subsets, each representing half of the lung, and only the subset with annotations was utilized for training and testing.This led to a final dataset consisting of 106 subjects. In addition, we also labeled the lung areas for two purposes. On one hand, this helps distinguish pulmonary arteries and veins inside and outside the lung.