Federated Learning for Brain Tumor Segmentation Using MRI and Transformers

Sahil Nalawade, Chandan Ganesh, Ben Wagner, Divya Reddy, Yudhajit Das, Fang F. Yu, Baowei Fei, Ananth J. Madhuranthakam, Joseph A. Maldjian

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This work focuses on training a deep learning network in a federated learning framework. The Federated Tumor Segmentation Challenge has 2 separate tasks. Task-1 was to design an aggregation logic for a given network, which is trained in a federated learning framework. Task-2 of the challenge was to train a network that is robust and generalizable in a federated testing environment. 341 subjects were used for training both tasks of the challenge. This data was distributed across 17 collaborators, which were then used to train an individual network for each collaborator. A new weight aggregation logic was developed. The network weights in this logic were determined based on the average validation dice scores of each collaborator. A concise model was obtained using the developed weighted aggregation logic. The Dice scores for task-1 on the validation dataset for whole tumor, tumor core, and enhancing tumor were 0.767, 0.612, and 0.628 respectively. The Dice scores for task-2 on the validation dataset for whole tumor, tumor core, and enhancing tumor were 0.874, 0.773, and 0.721 respectively.

Original languageEnglish (US)
Title of host publicationBrainlesion
Subtitle of host publicationGlioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 7th International Workshop, BrainLes 2021, Held in Conjunction with MICCAI 2021, Revised Selected Papers
EditorsAlessandro Crimi, Spyridon Bakas
PublisherSpringer Science and Business Media Deutschland GmbH
Pages444-454
Number of pages11
ISBN (Print)9783031090011
DOIs
StatePublished - 2022
Event7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021 - Virtual, Online
Duration: Sep 27 2021Sep 27 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12963 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference7th International Brain Lesion Workshop, BrainLes 2021, held in conjunction with the Medical Image Computing and Computer Assisted Intervention, MICCAI 2021
CityVirtual, Online
Period9/27/219/27/21

Keywords

  • Brain tumor
  • Convolutional neural network
  • Deep learning
  • Federated learning
  • Segmentation
  • Transformers

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Federated Learning for Brain Tumor Segmentation Using MRI and Transformers'. Together they form a unique fingerprint.

Cite this