Unsupervised Detection of Lung Nodules in Chest Radiography Using Generative Adversarial Networks

TL;DR

This paper introduces P-AnoGAN, an unsupervised deep learning model that detects lung nodules in chest X-rays with high accuracy, using only healthy images for training and demonstrating strong external validation results.

Contribution

The study presents P-AnoGAN, a novel unsupervised anomaly detection method utilizing a progressive GAN and encoder-decoder pipeline for lung nodule detection in radiographs.

Findings

Achieved ROC-AUC of 91.17% on validation data.

Achieved ROC-AUC of 87.89% on test data.

Demonstrated effectiveness of unsupervised methods in lung nodule detection.

Abstract

Lung nodules are commonly missed in chest radiographs. We propose and evaluate P-AnoGAN, an unsupervised anomaly detection approach for lung nodules in radiographs. P-AnoGAN modifies the fast anomaly detection generative adversarial network (f-AnoGAN) by utilizing a progressive GAN and a convolutional encoder-decoder-encoder pipeline. Model training uses only unlabelled healthy lung patches extracted from the Indiana University Chest X-Ray Collection. External validation and testing are performed using healthy and unhealthy patches extracted from the ChestX-ray14 and Japanese Society for Radiological Technology datasets, respectively. Our model robustly identifies patches containing lung nodules in external validation and test data with ROC-AUC of 91.17% and 87.89%, respectively. These results show unsupervised methods may be useful in challenging tasks such as lung nodule detection in radiographs.