Pulmonary Nodule Segmentation Method of CT Images Based on 3D-FCN.

In this work, we use a 3D Fully Convolutional Network (FCN) architecture for pulmonary nodule segmentation. Our method integrates FCN and Conditional Random Field(CRF) into an end-to-end network. Using this approach, the spatial features of CT image series can be better utilized to obtain the three-dimensional global features of pulmonary nodules according to the context. The model includes pulmonary nodule segmentation and classification recognition and the noise is reduced by effective image preprocessing. We achieved competitive results during the testing phase of the LIDC/IDRI dataset for segmentation and detection with sensitivity of 0.918 using 3D-FCN and VGG19.