Fully Automated Lung Lobe Segmentation in Volumetric Chest CT with 3D U-Net: Validation with Intra- and Extra-Datasets

Jongha Park, Jihye Yun, Namkug Kim, Beomhee Park, Yongwon Cho, Hee Jun Park, Mijeong Song, Minho Lee, Joon Beom Seo
Journal of Digital Imaging, May 2019
Related Product


Date Published
2019. 05

This study developed and validated a fully automated lung lobe segmentation method with 3D U-Net, using chest CT scans of 196 normal and mild-to-moderate COPD patients from three centers. Manual correction by a thoracic radiologist served as the gold standard for comparison. The deep learning method showed high accuracy in both internal and external validation when compared to the gold standards using metrics such as Dice similarity coefficient (DSC), Jaccard similarity coefficient (JSC), mean surface distance (MSD), and Hausdorff surface distance (HSD). The segmentation method was also time-efficient. Despite challenges in developing a robust automatic lung lobe segmentation method, this deep learning–based 3D U-Net method demonstrated promising accuracy and computational time, and could potentially be adapted for severe lung diseases in clinical workflows. The algorithm used for this method has been commercialized in AVIEW by Coreline Soft.


Please leave your inquiry if you have any questions regarding our products, recruitment, investment, or any other matters.

Contact us