TwoPath U-Net for automatic brain tumor segmentation from multimodal MRI data

Keerati Kaewrak, John Soraghan, Gaetano Di Caterina, Derek Grose

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

Abstract

A novel encoder-decoder deep learning network called TwoPath U-Net for multi-class automatic brain tumor segmentation task is presented. The network uses cascaded local and global feature extraction paths in the down-sampling path of the network which allows the network to learn different aspects of both the low-level feature and high-level features. The proposed network architecture using a full image and patches input technique was used on the BraTS2020 training dataset. We tested the network performance using the BraTS2019 validation dataset and obtained the mean dice score of 0.76, 0.64, and 0.58 and the Hausdorff distance 95% of 25.05, 32.83, and 37.57 for the whole tumor, tumor core and enhancing tumor regions.
Original languageEnglish
Title of host publicationBrainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries
Subtitle of host publication6th International Workshop, BrainLes 2020, Held in Conjunction with MICCAI 2020, Lima, Peru, October 4, 2020, Revised Selected Papers, Part II
EditorsAlessandro Crimi, Spyridon Bakas
Place of PublicationCham, Switzerland
PublisherSpringer International Publishing AG
Number of pages10
Volume12659
Edition1
ISBN (Electronic)978-3-030-72087-2
ISBN (Print)978-3-030-72086-5
Publication statusAccepted/In press - 5 Mar 2021
Event6th International Workshop, BrainLes 2020, Held in Conjunction with MICCAI 2020 - Online
Duration: 4 Oct 20204 Oct 2020
http://www.brainlesion-workshop.org/

Conference

Conference6th International Workshop, BrainLes 2020, Held in Conjunction with MICCAI 2020
Abbreviated titleMICCAI BrainLes 2020
Period4/10/204/10/20
Internet address

Keywords

  • brain tumor
  • deep learning
  • segmentation
  • TwoPath U-Net
  • multimodal MRI

Fingerprint Dive into the research topics of 'TwoPath U-Net for automatic brain tumor segmentation from multimodal MRI data'. Together they form a unique fingerprint.

Cite this