Accurate extraction of airways from computed tomography (CT) images of the lungs is essential for assessing pulmonary ventilation function and diagnosing respiratory diseases. Traditional airway segmentation methods rely heavily on manual interaction, limiting segmentation accuracy. Deep learning has been widely applied in medical image processing, especially in lung nodule detection and benign/malignant diagnosis. However, its application to airway segmentation in lung CT images faces challenges due to image noise and varying tissue densities, making it difficult to segment fine airways. Original CT images contain non-relevant regions such as bones and beds, increasing processing overhead and error rates. To address these issues, this paper proposes an Attention-Unet-based airway segmentation method using a stepwise processing strategy to enhance structural representation. Experimental results demonstrate that applying the Attention-Unet network to airway segmentation in lung CT images significantly improves segmentation speed and accuracy while effectively reducing leakage.
Merwood et al. (Tue,) studied this question.