多标签胸部X射线分类的深度学习方法比较.pdf

The increased availability of labeled X-ray image archives(e.g. ChestX-ray14 dataset) hastriggereda growing interest in deep learning techniques. To provide better insight into the different approaches, and their applicationsto chest X-ray classification, we investigate a powerful network architecturein detail: the ResNet-50. Building on prior workin this domain, we consider transfer learning with and without fine-tuning as well as the training of a dedicated X-ray network from scratch. To leveragethe high spatial resolution of X-ray data, we also include an extended ResNet-50 architecture, andanetwork integrating non-image data(patient age, gender and acquisition type) in the classification process. In a concluding experiment, we also investigate multiple ResNet depths(i.e. ResNet-38and ResNet-101). In a systematic evaluation, using 5-fold re-sampling and a multi-label loss function, we comparethe performance of the different approachesfor pathology classification by ROC statistics andanalyze differences between the classifiers using rank correlation. Overall, we observe a considerable spread in the achieved performance and conclude thatthe X-ray-specific ResNet-38, integrating non-image data yields the best overall results. Furthermore, class activation maps are used to understand the classification process, and a detailed analysis of the impact of non-image featuresisprovided.