In-depth analysis of medical data through machine learning to achieve disease prediction is beneficial to the early detection and treatment of diseases. However, medical data involves mass patient privacy, and datasets of different medical institutions cannot be directly shared due to privacy protection. So medical data often exists in the form of data islands, which makes it difficult for most existing prediction models to complete disease prediction. In this paper, a digital twin assisted efficient clustering Federated Learning (FL) algorithm for disease prediction is proposed. It can break data islands to predict diseases on the premise of privacy security. Firstly, we design an efficient clustering Federated Learning with Client Selection (FLCS) protocol based on heterogeneity and contribution to improve the training efficiency and prediction accuracy. Secondly, we use digital twin to assist the FLCS protocol to carry out large-scale prediction. In addition, the shapley value introduced in the calculation of client contribution makes the model interpretable and enhances the reliability of prediction results. Finally, the evaluation results show that compared with the common prediction models and FedAvg algorithm, the FLCS protocol assisted by digital twin has better efficiency and accuracy in binary classification prediction.

Projecting the original data into a low-dimensional target space for feature extraction is a common method. Recently, presentation-based approaches have been widely concerned and many feature extraction algorithms based on this have been proposed. However, in the process of acquiring real data, the pollution of complex noise cannot always be avoided, which will greatly increase the difficulty of feature extraction and even lead to failed feature extraction results. Thus, a robust image feature extraction model based on Orthogonal Low-Rank Projection Learning (OLRPL) is proposed, in which the introduction of orthogonal matrix can encourage the preservation of the main components of the sample. Particularly, the row sparsity constraint introduced on the projection matrix can encourage the features to be more compact, discriminative and interpretable. In particular, the Weighted Truncated Schattenp-norm (WTSN) is proposed to better solve the optimization problem of low-rank constraints. At the same time, the correntropy is applied in OLRPL to suppress the complex noise in the data and thus improve the robustness of the model. Finally, we specially design a robust classification loss function so that our model can be fitted the supervised scene effectively. Experiments on five general databases have proved that OLRPL has better effectiveness and robustness than existing advanced methods.