Fig. 4. Test accuracy of different methods on the pneumonia detection task.

The error bar represents the mean ± standard deviation of five repetitions. The results show that the proposed Selective-FD method achieves the best performance, and the accuracy gain is more significant when using hard labels to transfer knowledge. Specifically, some baselines perform even worse than the independent learning scheme. These results demonstrate that knowledge sharing among clients can mislead and negatively influence local training. The NIH Chest X-ray Dataset⁵³ and RSNA-STR Pneumonia Detection Challenge image datasets⁵⁴ are used for model training and testing. NIH Chest X-ray Dataset is available at https://nihcc.app.box.com/v/ChestXray-NIHCC, which is provided by the NIH Clinical Center. The RSNA-STR Pneumonia Detection Challenge image datasets is available at https://www.rsna.org/education/ai-resources-and-training/ai-image-challenge/RSNA-Pneumonia-Detection-Challenge-2018.