Self-Training with Entropy-Based Mixup for Low-Resource Chest X-ray Classification

Deep learning-based medical image analysis technology has been developed to the extent that it shows an accuracy surpassing ability of a human radiologist in some tasks. However, data labeling on images requires experts and great deal time expense. Moreover, usually have imbalanced distribution for each disease. In particular, multilabel classification, learning with small number labeled causes overfitting problems. The model easily overfits limited data, while still underfits large amount unlabeled data. this study, we propose method combines entropy-based Mixup self-training improve performance data-imbalanced chest X-ray classification. proposed is apply algorithm alleviate imbalance problem perform effectively utilizes iterating process by replacing teacher student model. Experimental results environment showed classification was improved combining self-training.