Hybrid optimisation for early Parkinson's disease detection in federated learning

Recently, a rapidly developing neurodegenerative disorder is Parkinson's disease (PD) which commonly affects the population of elder persons of individuals over 50 years. There still has been no medication for PD. Despite that, in the early stages, detecting PD is challenging. Thus, earlier detection is required to maximise the life of patients. In this research, the major intention is to project FedL_WFSA_ResneXt-DNN for identifying PD detection. Two entities are involved such as nodes and servers, whereas the models that exist in FL are training and global models. By acquiring the input image from the given dataset, PD detection takes place inside the local training model and the input image is pre-processed using a weighted median filter. The essential features are extracted and then, PD detection is performed by ResNeXt-DNN and the hyperparameters tuning is conducted utilising the proposed WFSA.