Privacy-first health research with federated learning

Sadilek, Adam; Liu, Luyang; Nguyen, Dung Tien; Kamruzzaman, Methun; Rader, Benjamin; Ingerman, Alex; Mellem, Stefan; Kairouz, Peter; Nsoesie, Elaine O.; MacFarlane, Jamie; Vullikanti, Anil; Marathe, Madhav V.; Eastham, P. R.; Brownstein, John S.; Howell, Michael D.; Hernandez, John

doi:10.1101/2020.12.22.20245407

Cited by 6 publications

(4 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other proposed solutions to improving data sharing include promising new technologies, such as synthetic data or federated learning [ 53 – 55 ], which have been suggested to potentially help researchers publicly sharing health data while better managing the risk of deidentification. However, linkage risk will always remain a concern as even releasing summary statistics alone constitutes a certain loss of privacy for the contributing data sources in terms of differential privacy [ 28 , 56 ].…”

Section: Potential Solutionsmentioning

confidence: 99%

Global healthcare fairness: We should be sharing more, not less, data

et al. 2022

View full text Add to dashboard Cite

The availability of large, deidentified health datasets has enabled significant innovation in using machine learning (ML) to better understand patients and their diseases. However, questions remain regarding the true privacy of this data, patient control over their data, and how we regulate data sharing in a way that that does not encumber progress or further potentiate biases for underrepresented populations. After reviewing the literature on potential reidentifications of patients in publicly available datasets, we argue that the cost—measured in terms of access to future medical innovations and clinical software—of slowing ML progress is too great to limit sharing data through large publicly available databases for concerns of imperfect data anonymization. This cost is especially great for developing countries where the barriers preventing inclusion in such databases will continue to rise, further excluding these populations and increasing existing biases that favor high-income countries. Preventing artificial intelligence’s progress towards precision medicine and sliding back to clinical practice dogma may pose a larger threat than concerns of potential patient reidentification within publicly available datasets. While the risk to patient privacy should be minimized, we believe this risk will never be zero, and society has to determine an acceptable risk threshold below which data sharing can occur—for the benefit of a global medical knowledge system.

show abstract

Section: Potential Solutionsmentioning

confidence: 99%

Global healthcare fairness: We should be sharing more, not less, data

et al. 2022

View full text Add to dashboard Cite

show abstract

“…However, experiences collected in one institution, or geographical region are seldom sufficient to solve diverse problems, especially for rare conditions. Yet, vast amount of data collected across institutions remain in silos where they are collected [14]- [16].…”

Section: Introductionmentioning

confidence: 99%

F-CBR: An Architecture for Federated Case-Based Reasoning

2022

View full text Add to dashboard Cite

Case-based reasoning (CBR) is a problem-solving methodology in artificial intelligence that attempts to solve new problems using past experiences known as cases. Experiences collected in a single case base from an institution or geographical region are seldom sufficient to solve diverse problems, especially in rare situations. Additionally, many institutions do not promote peer-to-peer (p2p) communication or encourage data sharing through such networks to retain autonomy. The paper proposes a federated CBR (F-CBR) architecture to address these challenges. F-CBR enables solving new problems based on similar cases from multiple autonomous CBR systems without p2p communication. We also designed an algorithm to minimize (irrelevant or unsolicited) data sharing in an F-CBR system. We extend the F-CBR design to support institutions with organizational or geographical hierarchies. The F-CBR architecture was implemented and evaluated on two public datasets and a private real-world (non-specific musculoskeletal disorder patient) dataset. The findings demonstrate that the retrieval quality of F-CBR systems is comparable to or better than a single CBR system that persists all the cases on a centralized case base. F-CBR systems address data privacy by incorporating the data minimization principle. We foresee F-CBR as a viable real-world design that can aid in federating legacy CBR systems with minimal or no changes. The CBR systems used in this study are shared on GitHub to support reproducibility. INDEX TERMSCase-based reasoning, data minimization, data privacy, data silos, decision support systems, federated architecture, federated case-based reasoning. TABLE 1. Literature comparison table.

show abstract

“…Federated learning (FL) has emerged as a promising distributed learning approach, where the parties keep their raw data on their premises and exchange intermediate model parameters 9,10 . This approach has enabled collaborative learning for several medical applications, and it has been shown that FL performs comparably to centralized training on medical datasets [11][12][13] . Recently, the concept of swarm learning (SL) has been proposed; it enables decentralized machine learning for precision medicine.…”

Section: Introductionmentioning

confidence: 99%

Privacy-Preserving Federated Neural Network Learning for Disease-Associated Cell Classification

Sav

Bossuat²,

Troncoso-Pastoriza³

et al. 2022

Preprint

View full text Add to dashboard Cite

Training accurate and robust machine learning models requires a large amount of data that is usually scattered across data-silos. Sharing or centralizing the data of different healthcare institutions is, however, unfeasible or prohibitively difficult due to privacy regulations. In this work, we address this problem by using a novel privacy-preserving federated learning-based approach, PriCell, for complex machine learning models such as convolutional neural networks. PriCell relies on multiparty homomorphic encryption and enables the collaborative training of encrypted neural networks with multiple healthcare institutions. We preserve the confidentiality of each institutions' input data, of any intermediate values, and of the trained model parameters. We efficiently replicate the training of a published state-of-the-art convolutional neural network architecture in a decentralized and privacy-preserving manner. Our solution achieves an accuracy comparable to the one obtained with the centralized solution, with an improvement of at least one-order-of-magnitude in execution time with respect to prior secure solutions. Our work guarantees patient privacy and ensures data utility for efficient multi-center studies involving complex healthcare data.

show abstract

Privacy-first health research with federated learning

Cited by 6 publications

References 38 publications

Global healthcare fairness: We should be sharing more, not less, data

Global healthcare fairness: We should be sharing more, not less, data

F-CBR: An Architecture for Federated Case-Based Reasoning

Privacy-Preserving Federated Neural Network Learning for Disease-Associated Cell Classification

Contact Info

Product

Resources

About