Segmentation of the Airway Tree From Chest CT Using Tiny Atrous Convolutional Network

The airway tree is one of the most important part in human respiratory system. Airway segmentation plays a crucial role pulmonary disease diagnosis, localization and surgical navigation. We propose novel method to improve thoracic computed tomography(CT) using deep learning. In order take into account multi-scale changes achieve accurate segmentation, we design an end-to-end Tiny Atrous Convolutional Network (TACNet) based on 3D convolution neural network. view difficulty classification due numerous branches airway, two evaluation factors, namely, angle bifurcation buffer length are designed, which used for by combining centerline extraction. train TACNet inspirator CT scans with ground truth, generated clinicians evaluate our own clinical data sets EXACT'09 sets. Compared state-of-the-art algorithms, proposed algorithm this paper very competitive results 20 test datasets challenge. experimental show that has high robustness advantages regardless or classification. challenge, average detection rate best public literatures.