Deep Learning Approach for Auto-Detecting Idiopathic Pulmonary Fibrosis Prediction

摘要

In the field of computer vision, Convolutional Neural Network has been the most mainstream method and has shown excellent performance in medical images. Among Convolutional Neural Networks, U-Net and DenseNet have demonstrated outstanding and robust performance in image recognition and image segmentation, respectively. In this paper, we proposed a neural network with DenseNet as the Encoder and Unet as the Decoder for lung image segmentation and feature extraction. With this neural network, we extracted features from patients' CT Scan images and combined them with patients' clinical records to predict lung function trends in the future. This predictive value will provide significant help in determining whether the patient has Idiopathic Pulmonary Fibrosis, which is the purpose of our study.