Searching for Pneumothorax in X-Ray Images Using Autoencoded Deep Features
Fast diagnosis and treatment of pneumothorax, a collapsed or dropped lung, is crucial to avoid fatalities. Pneumothorax is typically detected on a chest X-ray image through visual inspection by experienced radiologists. However, the detection rate is quite low. Therefore, there is a strong need for automated detection systems to assist radiologists. Despite the high accuracy levels generally reported for deep learning classifiers in many applications, they may not be useful in clinical practice due to the lack of large number of high-quality labelled images as well as a lack of interpretation possibility. Alternatively, searching in the archive of past cases to find matching images may serve as a 'virtual second opinion' through accessing the metadata of matched evidently diagnosed cases. To use image search as a triaging/diagnosis tool, all chest X-ray images must first be tagged with identifiers, i.e., deep features. Then, given a query chest X-ray image, the majority vote among the top k retrieved images can provide a more explainable output. While image search can be clinically more viable, its detection performance needs to be investigated at a scale closer to real-world practice. We combined 3 public datasets to assemble a repository with more than 550,000 chest X-ray images. We developed the Autoencoding Thorax Net (short AutoThorax-Net) for image search in chest radiographs compressing three inputs: the left chest side, the flipped right side, and the entire chest image. Experimental results show that image search based on AutoThorax-Net features can achieve high identification rates providing a path towards real-world deployment. We achieved 92% AUC accuracy for a semi-automated search in 194,608 images (pneumothorax and normal) and 82% AUC accuracy for fully automated search in 551,383 images (normal, pneumothorax and many other chest diseases).