Federated Domain Adaptation via Transformer for Multi-Site Alzheimer's Disease Diagnosis.

In multi-site studies of Alzheimer's disease (AD), the difference of data in multi-site datasets leads to the degraded performance of models in the target sites. The traditional domain adaptation method requires sharing data from both source and target domains, which will lead to data privacy issue. To solve it, federated learning is adopted as it can allow models to be trained with multi-site data in a privacy-protected manner. In this paper, we propose a multi-site federated domain adaptation framework via Transformer (FedDAvT), which not only protects data privacy, but also eliminates data heterogeneity. The Transformer network is used as the backbone network to extract the correlation between the multi-template region of interest features, which can capture the brain abundant information. The self-attention maps in the source and target domains are aligned by applying mean squared error for subdomain adaptation. Finally, we evaluate our method on the multi-site databases based on three AD datasets. The experimental results show that the proposed FedDAvT is quite effective, achieving accuracy rates of 88.75%, 69.51%, and 69.88% on the AD vs. NC, MCI vs. NC, and AD vs. MCI two-way classification tasks, respectively.

Prevalence and patterns of multimorbidity in chronic diseases in Guangzhou, China: a data mining study in the residents' health records system among 31 708 community-dwelling elderly people.

OBJECTIVES: Examination of the prevalence, influence factors and patterns of multimorbidity among the elderly people in Guangzhou, China. DESIGN: Cross-sectional study. PARTICIPANTS: 31 708 community-dwelling elderly people aged 65 and over. PRIMARY AND SECONDARY OUTCOME MEASURES: Prevalence, influence factors and patterns of multimorbidity in seven chronic conditions among the participants. A multistage, stratified random sampling was adopted for selection of health records in the residents' health records system of Guangzhou. Data mining by association rule mining analysis was used to explore the correlations and multimorbidity patterns between seven chronic diseases. RESULTS: The prevalence of morbidity was 55.0% (95% CI 40.1% to 60.1%) and the multimorbidity was 15.2% (95% CI 12.4% to 18.4%) among the participants. Elderly, women, higher education level, being single, living in urban areas and having medical insurance were more likely to have chronic diseases and multimorbidity. Data mining by association rule mining analysis reveals patterns of multimorbidity among the participants, including coexistence of hypertension and diabetes (support: 12.5%, confidence: 17.6%), hypertension and coronary heart disease (support: 4.4%, confidence: 5.7%), diabetes and coronary heart disease (support: 1.6%, confidence: 5.7%), diabetes, coronary heart disease and hypertension (support: 1.4%, confidence: 4.4%). CONCLUSIONS: A high prevalence of morbidity (especially on hypertension and diabetes) and a relatively low multimorbidity of chronic diseases exist in elderly people. Data mining of residents' health records will help for strengthening the management of residents' health records in community health service centres of Guangzhou, China.