Pulmonary bronchial segmentation scheme based on 3D UX-Net

Le résumé fourni par la source

In order to achieve accurate segmentation of lung bronchus and obtain the inner diameter of the bronchus, a new method of lung bronchus segmentation design and post-processing optimization based on 3D UX-Net model was proposed. Firstly, 3D UX-Net model was used to train bronchial data. Secondly, a new post-processing process to improve segmentation accuracy is combined with region selection, threshold adjustment and maximum connected domain analysis. Finally, a comparison experiment with other algorithm models is conducted on the open data set. The experimental results showed that the Dice coefficient evaluation index based on the 3D UX-Net model was improved from 0.83 to 0.85, and the error of the published and clinical data sets on the inner diameter of the bronchus was within 0.2mm. It was concluded that the model and post-processing method proposed in this study could segment the bronchus quickly and accurately.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.