FV-Seg-Net: Fully Volumetric Network for Accurate Segmentation of COVID-19 Lesions from Chest CT Scans

Mohamed Abdel-Basset, Hossam Hawash, Victor Chang

Research output: Contribution to journalArticlepeer-review

Abstract

Current pneumonia segmentation approaches lack precision on small infection areas and operate by partitioning the CT volumes into 2D slices or 3D patches, leading to the loss of contextual information. We propose an improved fully volumetric segmentation network, called FV-SEG-Net, that effectively exploits the local and global spatial information and enables the entire CT volume processing. The encoder network is implemented with 3D ResNeXt. The decoder is designed using a computationally efficient recalibrated anisotropic convolution (RAC) module to acquire the 3D semantic representation of the CT volumes with anisotropic resolution. To avoid losing information during down-sampling, we reconstruct the skip- connection using a multi-level multi-scale pyramid aggregation (MPA) module and ensure more effective context fusion that improves the reconstruction decoder capability. Empirical investigations demonstrate that FV-SEG-Net has an excellent performance in segmenting COVID-19 lesions with a Dice score of 78.58% and a surface-Dice score of 80.1% outperforming current cutting-edge approaches.
Original languageEnglish
JournalIEEE Transactions on Industrial Informatics
DOIs
Publication statusE-pub ahead of print - 16 Sep 2022

Bibliographical note

© 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Keywords

  • 3D CT scans
  • Anisotropic Convolution
  • COVID-19
  • Computed tomography
  • Convolution
  • Deep Learning
  • Image segmentation
  • Lesions
  • Medical S upply chain
  • Solid modeling
  • Three-dimensional displays
  • Volumetric S egmentation

Fingerprint

Dive into the research topics of 'FV-Seg-Net: Fully Volumetric Network for Accurate Segmentation of COVID-19 Lesions from Chest CT Scans'. Together they form a unique fingerprint.

Cite this