Systematic Review of Privacy-Preserving Distributed Machine Learning From Federated Databases in Health Care

Abstract

Big data for health care is one of the potential solutions to deal with the numerous challenges of health care, such as rising cost, aging population, precision medicine, universal health coverage, and the increase of noncommunicable diseases. However, data centralization for big data raises privacy and regulatory concerns. Covered topics include (1) an introduction to privacy of patient data and distributed learning as a potential solution to preserving these data, a description of the legal context for patient data research, and a definition of machine/deep learning concepts; (2) a presentation of the adopted review protocol; (3) a presentation of the search results; and (4) a discussion of the findings, limitations of the review, and future perspectives. Distributed learning from federated databases makes data centralization unnecessary. Distributed algorithms iteratively analyze separate databases, essentially sharing research questions and answers between databases instead of sharing the data. In other words, one can learn from separate and isolated datasets without patient data ever leaving the individual clinical institutes. Distributed learning promises great potential to facilitate big data for medical application, in particular for international consortiums. Our purpose is to review the major implementations of distributed learning in health care.

Development of a Clinical Decision Support System for Severity Risk Prediction and Triage of COVID-19 Patients at Hospital Admission: an International Multicenter Study

“…Thirdly, the threshold to go to the hospital and hospitalisation management can vary from country to country, and we are also aware that RNA viruses can mutate rapidly and could have an impact of the performance of the models. We therefore propose that those models should be continuously updated to achieve a better performance for example using privacy preserving distributed learning approaches [ 32 , 33 ]. Fourthly, the CT features used for this study are semantic features from the first CT scan, and radiomics or deep learning approaches may improve its prognostic performance, and follow-up CT scans may yield more information.…”

Development of a Clinical Decision Support System for Severity Risk Prediction and Triage of COVID-19 Patients at Hospital Admission: an International Multicenter Study

“…We therefore propose that those models should be continuously updated for example using privacy-preserving distributed learning approaches. 29,30 Fourth, the CT features used for this study are semantic features from the first CT scan, and quantitative features automatically extracted from CT images using radiomics or deep learning approaches may improve its prognostic performance, and follow-up CT scan may yield more information. Finally, there is also the fundamental weakness of nomograms, which do not give a confidence interval to the final output.…”

Blockchain for Privacy Preserving and Trustworthy Distributed Machine Learning in Multicentric Medical Imaging (C-DistriM)

“…Similar works have been demonstrated by Kuo et al [45], [46] leveraging blockchain using Logistic Regression machine learning models. While these pipelines [13,32,33] overcome the risk of exposing the model weights, Lugan et al [47] proposed to train distributed learning models on encrypted data, preventing any exposure of local weights. Nevertheless, when implementing deep learning and encrypting model weights, model design requires careful consideration as aspects such as the CNN activation functions must be adapted [48].…”

“…This approach, proposed in 2013, is known as distributed learning (federated learning) [11], [12]. Distributed learning -a fusion of machine learning and distributed computing -allows machine learning models to be trained on multiple siloed datasets without the need for patient data to leave the firewalls of each database [13]. Distributed learning preserves privacy by design, by sharing model weights for subsequent training cycles instead of privacy sensitive data.…”