Multi-institutional Deep Learning Modeling Without Sharing Patient Data: A Feasibility Study on Brain Tumor Segmentation

Abstract

Deep learning models for semantic segmentation of images require large amounts of data. In the medical imaging domain, acquiring sufficient data is a significant challenge. Labeling medical image data requires expert knowledge. Collaboration between institutions could address this challenge, but sharing medical data to a centralized location faces various legal, privacy, technical, and data-ownership challenges, especially among international institutions. In this study, we introduce the first use of federated learning for multi-institutional collaboration, enabling deep learning modeling without sharing patient data. Our quantitative results demonstrate that the performance of federated semantic segmentation models (Dice=0.852) on multimodal brain scans is similar to that of models trained by sharing data (Dice=0.862). We compare federated learning with two alternative collaborative learning methods and find that they fail to match the performance of federated learning.

Segmentation and Classification in Digital Pathology for Glioma Research: Challenges and Deep Learning Approaches

“…One of the major challenges in machine learning analysis of biomedical imaging data is the lack of large curated and annotated training datasets, primarily because of time effort and domain expertise required for manual segmentations and classifications of tissue regions and micro-anatomic structures, such as nuclei and cells, as well as because of privacy and ownership concerns of source datasets. Some initial studies in the field of distributed learning in medicine attempted to address the data privacy and ownership challenge (Chang et al, 2018b;Sheller et al, 2018). These approaches need more investigation and adoption to facilitate collaboration across multiple medical institutions.…”

Privacy-preserving quality control of neuroimaging datasets in federated environment

“…The Intel corporation has started a collaboration with the University of Pennsylvania and 19 other institutions to advance real world medical research using the federated learning. They showed that a deep learning model which is trained by traditional approach on pooled data, can be trained up to 99% of the pooled accuracy using the federated learning approach [53].…”

Brain Tumor Analysis Empowered with Deep Learning: A Review, Taxonomy, and Future Challenges

“…The federated learning approach is used for semantic segmentation without sharing patient data by the multi-institutional collaboration. Federated learning provides better accuracy for semantic segmentation, with respect to the model that is trained on sharing data [145].…”