[Paper Review] Federated Learning for Healthcare Informatics

Jie Xu, Benjamin S. Glicksberg+4 more|arXiv (Cornell University)|Nov 13, 2019

Privacy-Preserving Technologies in Data164 references65 citations

TL;DR

This paper surveys federated learning for healthcare informatics, detailing statistical, system, and privacy challenges, methods to address them, and healthcare applications.

ABSTRACT

With the rapid development of computer software and hardware technologies, more and more healthcare data are becoming readily available from clinical institutions, patients, insurance companies and pharmaceutical industries, among others. This access provides an unprecedented opportunity for data science technologies to derive data-driven insights and improve the quality of care delivery. Healthcare data, however, are usually fragmented and private making it difficult to generate robust results across populations. For example, different hospitals own the electronic health records (EHR) of different patient populations and these records are difficult to share across hospitals because of their sensitive nature. This creates a big barrier for developing effective analytical approaches that are generalizable, which need diverse, "big data". Federated learning, a mechanism of training a shared global model with a central server while keeping all the sensitive data in local institutions where the data belong, provides great promise to connect the fragmented healthcare data sources with privacy-preservation. The goal of this survey is to provide a review for federated learning technologies, particularly within the biomedical space. In particular, we summarize the general solutions to the statistical challenges, system challenges and privacy issues in federated learning, and point out the implications and potentials in healthcare.

Motivation & Objective

• Motivate the need to learn from fragmented and private healthcare data without sharing raw data.
• Summarize federated learning foundations and its relevance to healthcare data like EHRs and wearables.
• Categorize and review statistical, system, and privacy challenges and proposed solutions.
• Illustrate healthcare-specific applications and representative methods.
• Discuss open questions and future directions for FL in healthcare.

Proposed method

• Provide a formal overview of federated learning and its objective to train a global model from decentralized data.
• Summarize and categorize challenges into statistical, communication, and privacy/security aspects.
• Discuss consensus vs. pluralistic approaches to non-IID data distributions.
• Review methods to improve communication efficiency (model compression, client selection, updates reduction, peer-to-peer learning).
• Describe privacy techniques including secure multi-party computation and differential privacy, and their trade-offs.

Experimental results

Research questions

• RQ1 What are the main statistical, system, and privacy challenges in federated learning as applied to healthcare data?
• RQ2 What solutions and methods address non-IID data, communication bottlenecks, and privacy preservation in healthcare FL?
• RQ3 How has federated learning been applied to healthcare tasks such as EHR analysis, phenotyping, and mortality/prediction modeling?
• RQ4 What open questions and future directions exist for deploying FL in healthcare informatics?

Key findings

• Federated learning enables training across fragmented healthcare data while keeping data local, addressing privacy concerns.
• AFL and q-Fair Federated Learning are proposed approaches to handle non-IID distributions and fairness across devices.
• Privacy-preserving techniques include secure multi-party computation and differential privacy, with trade-offs in computational cost and accuracy.
• Applications in healthcare include patient similarity learning, phenotyping, representation learning from multimodal data, and mortality/prediction tasks; Table 1 summarizes representative works.
• Communication efficiency strategies are categorized into model compression, client selection, updates reduction, and peer-to-peer learning, addressing practical deployment challenges.