Segmentation and Classification Method of Pulmonary Nodule Based on Neural Network Model

摘要

In this paper, to realize automatic detection and classification of pulmonary nodules, U-Net neural network was used for nodular segmentation, and then the random forest algorithm was used to do the binary judgment judging whether the nodules were cancer. Based on the lung CT file and nodule annotation information in the mhd format of the LUNA2016 dataset, this paper extracted the ROI region of the lung through data preprocessing, and generated a mask based on the nodule position information, and then sent it into the U-Net model for nodule segmentation. According to the nodule information and classification files, the random forest algorithm was used to successfully classify and judge whether the nodules are cancerous.