Purpose
Deep learning (DL) models show promise in a wide variety of medical imaging tasks [1]. Training DL models, however, often requires significant amounts of data. To improve generalizability and overcome biases of individual institutions, extensive and diverse multi-institutional training datasets are desired [1]. The Federated Tumor Segmentation (FeTS, https://www.fets.ai) initiative, spearheaded by the University of Pennsylvania, describes the largest international federation of healthcare sites and a user-friendly software tool that enables training DL models for brain tumor segmentation. FeTS focuses on leveraging information gathered from...
Methods and materials
Histologically-confirmed cases of glioblastoma managed under the neurosurgical department at Flinders Medical Centre (FMC, South Australia) between 2010 and 2019 were identified. We contributed 110 cases to the FeTS training. Pre-operative baseline magnetic resonance imaging (MRI) scans comprising volumes acquired as T1-weighted, T1-weighted after Gadolinium contrast enhancement (T1-ce), T2-weighted, and T2-weighted-Fluid-Attenuated Inversion Recovery (FLAIR) were extracted from FMC’s imaging server. Pre-processing of the imaging data consisted of image registration to a standard atlas [3] and skull-stripping [4]. Initial inference performed by the FeTS platform produced...
Results
In a separate study, the initial phase of the federation was conducted with 11 collaborating sites, and it focused on addressing any potential network communication issues and with the segmentation task being focused solely on the tumor core region. The conclusion of the initial phase yielded an improvement in the DICE Similarity Coefficient for the tumor core of 11%, when comparing the performance of the consensus model with the model trained only with publicly available data. By conducting this study with much larger number of...
Conclusion
The FeTS initiative serves as an example for how to conduct multi-site and multi-national collaborative in DL for medical imaging. In addition to improving the performance and generalizability of DL models, a federated learning approach also overcomes privacy and legal challenges associated with sharing patient data. Successful development of the FeTS platform will enable a user-friendly brain tumor segmentation software tool without any requirement a computational background by the user.
References
Sheller, Micah J et al. “Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data.”Scientific reports vol. 10,1 12598. 28 Jul. 2020.
Sheller M, Reina G, Edwards B, Martin J, Bakas S. Multi-institutional Deep Learning Modeling Without Sharing Patient Data: A Feasibility Study on Brain Tumor Segmentation. Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. 2019;92-104.
Rohlfing T, Zahr N, Sullivan E, Pfefferbaum A. The SRI24 multichannel atlas of normal adult human brain structure. Human Brain Mapping. 2009;31(5):798-819.
Thakur S, Doshi J, Pati...