Abstract
Early and accurate diagnosis of interstitial lung diseases (ILDs) is crucialfor making treatment decisions, but can be challenging even for experiencedradiologists. The diagnostic procedure is based on the detection andrecognition of the different ILD pathologies in thoracic CT scans, yet theirmanifestation often appears similar. In this study, we propose the use of adeep purely convolutional neural network for the semantic segmentation of ILDpatterns, as the basic component of a computer aided diagnosis (CAD) system forILDs. The proposed CNN, which consists of convolutional layers with dilatedfilters, takes as input a lung CT image of arbitrary size and outputs thecorresponding label map. We trained and tested the network on a dataset of 172sparsely annotated CT scans, within a cross-validation scheme. The training wasperformed in an end-to-end and semi-supervised fashion, utilizing both labeledand non-labeled image regions. The experimental results show significantperformance improvement with respect to the state of the art.