Federated learning enables big data for rare cancer boundary detection

Abstract

Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.

Clinical Informatics Approaches to Facilitate Cancer Data Sharing

“…One of the largest demonstrations of federated learning at scale was published by Pati et al [27] With a group of 71 institutions across six continents the authors trained a successful auto-segmentation model for glioblastomas on brain MRIs. As expected, the authors found increased data improved overall model performance and made federated models more robust to potential data quality issues at individual institutions.…”

Democratizing clinical-genomic data: How federated platforms can promote benefits sharing in genomics

“…However, there also exists the potential for partial federation (I and II), when either compute access or data access are federated and compute or databases are distributed ( Table 1 ). This is distinct from federated learning, which has tackled this problem in the context of Machine Learning (ML) in healthcare—researchers can train machine algorithms collaboratively on dispersed data, including health records, without infringing on data governance legislations ( Mandl and Kohane, 2015 ; Stephens et al, 2015 ; De Fauw et al, 2018 ; Rieke et al, 2020 ; Xu et al, 2021 ; Pati et al, 2022 ).…”

Artificial Intelligence for Scientific Discovery at High-Performance Computing Scales

“…One way of alleviating this issue is with federated learning, whereby data storage is decentralized, and the only information transferred among workers in this massively distributed setting is model parameters. As a recent and perhaps largest example of federated learning for biological applications, Pati et al 22 demonstrated the need for highly generalizable AI models applied to large, diverse datasets. In this case, an aggregate, "consensus model" Another avenue for AI in chemistry that has seen a significant amount of interest is representation learning.…”